Note: Descriptions are shown in the official language in which they were submitted.
Persona-Based Conversational Interface Personalization Using
Social Network Preferences
TECHNICAL FIELD
[0001] The present disclosure relates to computer-implemented methods,
software, and systems for personalizing interactions within a conversational
interface
based on an input context and, in some instances, social network information
associated
with a user associated with the input.
BACKGROUND
[0002] Digital, or virtual, personal assistants such as AppleTM Inc.'s SinTM,
GoogleTM LLC's Google Assistant, AmazonTM Technologies Inc.'s Alexa,
MicrosoftTM
Corporation's CortanaTM, and others provide solutions for performing tasks or
services
associated with an individual. Such digital personal assistants can be used to
request and
perform various data exchanges, including transactions, social media
interactions, search
engine queries, and others. Additionally, similar functionality can be
incorporated into
web browsers and dedicated applications. Digital assistants may be one type of
conversational interface, where users can input a request or statement into
the
conversational interface and receive semantic output responsive to the
original input.
Conversational interfaces may be included in social networks, mobile
applications,
instant messaging platforms, websites, and other locations or applications.
Conversational interfaces may be referred to as or may be represented as chat
bots,
instant messaging (IM) bots, interactive agents, or any other suitable name or
representation.
[0003] Conversational interfaces are commonly integrated into dialog systems
of
these digital assistants or into specific applications or platforms, and can
be used to
engage in interactions from casual conversations to expert system analysis.
Conversational interfaces may accept inputs and/or output responses in various
formats,
including textual inputs and/or outputs, auditory inputs and/or outputs, video-
captured
inputs (e.g., via facial movement input), or video or other animated output.
Date Recue/Date Received 2020-09-20
SUMMARY
[0004] The present disclosure involves systems, software, and computer
implemented methods for personalizing interactions within a conversational
interface. A
first example system includes a communications module, at least one memory
storing
instructions and a repository of synonym tokens, and at least one hardware
processor
interoperably coupled with the at least one memory and the communications
module.
The repository of synonym tokens can include synonym tokens for association
with one
or more received inputs, where each of the synonym tokens associated with a
corresponding predefined lexical personality score. The instructions stored in
the at least
one memory can instruct the at least one hardware processor to perform various
operations. For example, the instructions can cause the at least one processor
to receive,
via the communications module, a first signal including a conversational input
received
via interactions with a conversational interface. The received conversational
input can be
analyze via a natural language processing engine to determine an intent of the
received
conversational input and a lexical personality score of the received
conversational input.
The determined intent and the determined lexical personality score can be
based on
characteristics included within the received conversational input. Next, a set
of response
content responsive to the determined intent of the received conversational
input can be
determined, where the response content includes a set of initial tokens
representing an
initial response to the received conversational input. A set of synonym tokens
associated
with at least some of the set of initial tokens can be identified, and, from
the identified set
of synonym tokens, at least one synonym token associated with a lexical
personality
score similar to the determined lexical personality score of the received
conversational
input can be determined. At least one token from the set of initial tokens
included in the
determined set of response content can be replaced with the at least one
determined
synonym token to generate a modified version of the set of response content.
In response
to the received first signal and via the communications module, a second
signal including
the modified version of the set of response content can be transmitted to a
device
associated with the received conversational input.
2
CA 2981261 2017-10-04
[0005] Implementations can optionally include one or more of the following
features.
[0006] In some instances, the received conversational input comprises a
semantic
query, wherein the intent of the conversational input is the semantic query.
[0007] In some instances, the determined lexical personality score of the
received
conversational input comprises at least a first score representing a relative
politeness of
the received conversational input and a second score representing a relative
formality of
the received conversational input.
[0008] In some of those instances, the determined lexical personality score of
the
received conversational input can comprise a third score representing an
identification of
at least one regional phrase included within the received conversational
input.
[0009] The received conversational input can, in some instances, comprise
textual
input received via the conversational interface.
[0010] In some instances, the received conversational input comprises audio
input
comprising a verbal statement received via the conversational interface. In
some of those
instances, the determined lexical personality score of the received
conversational input
can comprise a third score representing a determined accent of a speaker
associated with
the verbal statement. In those instances, determining the at least one synonym
token
associated with the lexical personality score similar to the determined
lexical personality
score of the received conversational input can includes determining at least
one synonym
token associated with the determined accent.
[0011] In some instances where the input comprises audio input, the determined
lexical personality score of the received conversational input can comprise a
third score
representing a verbal tone of a speaker associated with the verbal statement.
In those
instances, determining the at least one synonym token associated with the
lexical
personality score similar to the determined lexical personality score of the
received
conversational input can include determining at least one synonym token
associated with
the determined verbal tone of the speaker.
[0012] In some instances, each token in the set of initial tokens representing
the
initial response to the received conversational input can comprise a
particular phrase
3
CA 2981261 2017-10-04
within a responsive sentence represented by the set of response content. Each
of the
identified set of synonym tokens can then comprise a phrase having a similar
meaning to
at least one of the particular phrases in the responsive sentence.
[0013] In some instances, the modified version of the set of response content
corresponds to at least a formality and politeness level of the received
conversational
input.
[0014] A second example system includes a communications module, at least one
memory, and at least one hardware processor interoperably coupled with the at
least one
memory and the communications module. The at least one memory can store
instructions, a plurality of user profiles, and a repository of persona-
related contextual
content associated with a plurality of personas, where the persona-related
contextual
content for use in personalizing at least one response generated in response
to a
conversational contextual input. The instructions can instruct the at least
one hardware
processor to receive, via the communications module, a first signal including
a
conversational input received via interactions with a conversational
interface. The
conversational input is associated with a particular user profile, which is
itself associated
with a set of social network activity information. The received conversational
input is
analyzed via a natural language processing engine to determine an intent of
the
received conversational input and to determine a personality input type of the
received
conversational input. A persona response type associated with the determined
personality
input type is identified, and a set of response content responsive to the
determined intent
of the received conversational input is determined. A particular persona
associated with
the particular user profile is identified based on the set of social network
activity
information and corresponding to the identified persona response type is
identified, where
the particular persona is associated with a set of persona-related content.
The set of
response content is then modified using at least a portion of the persona-
related content to
generate a persona-associated response. A second signal is then transmitted
via the
communications module to a device associated with the particular user profile,
the second
signal including the persona-associated response for presentation in response
to the
received conversational input.
4
CA 2981261 2017-10-04
[0015] Implementations can optionally include one or more of the following
features.
[0016] In some instances, the determined personality input type is one of a
plurality of predefined personality input types, wherein each predefined
personality input
type is mapped to a persona response type.
[0017] In some instances, identifying a particular persona associated with the
particular user profile is further based on at least one of a current context
of the particular
user profile, a financial history of the particular user profile, and a
financial analysis
associated with the particular user profile.
[0018] In some instances, the determined intent associated with the received
conversational input is associated with a question, and wherein determining
the set of
response content responsive to the determined intent of the received
conversational input
comprises determining a responsive answer to the question. In those instances,
the
responsive answer to the question can be associated with a preferred action to
be
performed by the user associated with the particular user profile, wherein
identifying the
particular persona associated with the particular user profile based on the
associated set
of social network activity includes identifying the particular persona based
on the
preferred action to be performed by the user.
[0019] In some instances, the set of social network activity information is
stored
remotely from the particular user profile, wherein the particular user profile
is associated
with at least one social network account, and wherein the set of social
network activity is
accessed in response to the conversational input and prior to identifying the
particular
persona associated with the particular user profile.
[0020] In some instances, the set of social network activity information
identifies
at least one social network account followed by, liked, or subscribed to by
the particular
user profile, and identifying the particular persona associated with the
particular user
profile can comprise identifying a particular persona from the plurality of
personas
corresponding to the at least one social network account followed by, liked,
or subscribed
to by the particular user profile.
CA 2981261 2017-10-04
[0021] In alternative or additional instances, the set of social network
activity
information identifies at least one social network account with which the
particular user
profile has previously had a positive interaction, and identifying the
particular persona
associated with the particular user profile can comprise identifying a
particular persona
from the plurality of personas corresponding to the at least one social
network account
with which the particular user profile has had the positive interaction.
[0022] In some instances, for each of the personas, the persona-related
contextual
content includes a set of common phrases or words associated with the
particular persona.
Modifying the set of response content using at least a portion of the persona-
related
contextual content to generate a persona-associated response for the
particular user can
include incorporating at least one common phrase or word associated with the
identified
particular persona into the set of response content. In some instances,
incorporating the
at least one common phrase or word associated with the identified particular
persona into
the set of response content comprises replacing at a portion of the set of
response content
with at least one common phrase or word associated with the identified
particular
= persona.
[0023] In some instances, for at least some of the personas, the persona-
related
contextual content includes a voice associated with the persona, and wherein
modifying
the set of response content using at least a portion of the persona-related
contextual
content to generate a persona-associated response for the particular user
comprises
generating an audio file for use in presenting the set of response content
spoken in the
voice associated with the identified particular persona.
[0024] Similar operations and processes may be performed in a system
comprising at least one process and a memory communicatively coupled to the at
least
one processor where the memory stores instructions that when executed cause
the at least
one processor to perform the operations. Further, a non-transitory computer-
readable
medium storing instructions which, when executed, cause at least one processor
to
perform the operations may also be contemplated. Additionally, similar
operations can
be associated with or provided as computer implemented software embodied on
tangible,
non-transitory media that processes and transforms the respective data, some
or all of the
6
CA 2981261 2017-10-04
aspects may be computer implemented methods or further included in respective
systems
or other devices for performing this described functionality. The details of
these and
other aspccts and embodiments of the present disclosure are set forth in the
accompanying drawings and the description below. Other features, objects, and
advantages of the disclosure will be apparent from the description and
drawings, and
from the claims.
DESCRIPTION OF DRAWINGS
[0025] FIG. 1 is a block diagram illustrating an example system for
personalizing
interactions with a conversational interface.
[0026] FIGS. 2A-B is an illustration of a data and control flow of example
interactions performed by a system performing personalization operations with
a
conversational interface related to an analysis of the lexical personality of
an input, where
the responsive output is provided with an output lexical personality based on
the input
lexical personality.
[0027] FIG. 3 is a flow chart of an example method performed at a system
associated with a conversational interface to identify a first lexical
personality of an input
and provide a response to the input applying a second lexical personality
based on the
identified first lexical personality.
[0028] FIG. 4 is an illustration of a data and control flow of example
interactions
performed by a system performing persona-based personalization operations for
a
conversational interface based on an analysis of the lexical personality of an
input and a
corresponding user profile, generating a responsive output associated with the
input, and
modifying or transforming the responsive output to a persona-specific
responsive output
based on the corresponding user profile associated with the received input.
[0029] FIG. 5 is a flow chart of an example method performed at a system
associated with a conversational interface to transform a responsive output
into a
persona-specific responsive output based on the corresponding user profile.
7
CA 2981261 2017-10-04
DETAILED DESCRIPTION
[0030] The present disclosure describes various tools and techniques
associated
with personalizing a responsive output generated by a conversational interface
based on a
personality of the received input to which the output is responsive.
Conversational
interfaces allow users to interact with a virtual person or system that can
interpret an
intent associated with input provided by the user and determine a suitable
response.
[0031] When building a conversational interface, the personality is an
important
aspect of the design. Generally, building rapport with individuals is a
difficult task, but
one that can improve the level of engagement and responsive action that the
conversational interface may elicit from a user, and can consequently improve
the trust
that the user may have with such an interface, as well as with the entity or
organization
associated with the interface. A difficulty of current conversational
interfaces is that most
are associated with a single personality. In using a single personality for
all types of
feedback, such conversational interfaces fail to provide a dynamic and
believable
experience for users. The present disclosure describes several solutions for
enhancing
responsive content and dynamically adjusting the output of the conversational
interface to
match the lexical personality and/or preferences of the user. In doing so, the
present
solution enhances the conversational interface experience by transforming
interactions
into a realistic and dynamic experience.
[0032] In a first solution, and to enhance the interactions with the
conversational
interface, a method and system for identifying the personality of a
conversational
interface user based on measured characteristics of their conversational
pattern is
provided. The measured characteristics can include, but are not limited to, a
particular
formality, politeness, colloquial terminology, sarcasm, and emotional state of
received
input. Once the characteristics of the user's input are inferred from the
conversational
patterns of the user, those characteristics can be used in the natural
language generation
process to identify and apply a corresponding lexical output personality to be
applied to a
particular response prior to transmitting the response back to the user.
[0033] The detection and measuring of the input's characteristics can be
quantified through a scoring mechanism to identify particular measured
components of
8
CA 2981261 2017-10-04
the input. For instance, a formality score and a politeness score can be
quantified by a
natural language processing engine associated with the conversational
interface. Any
suitable scoring and natural language processing techniques can be used to
analyze and
determine the measured characteristics, and any suitable individual
characteristics can be
measured.
[0034] Using any suitable component, engine, or method, an intent associated
with the input can be determined. Once an intent of the input is determined, a
set of
responsive content can be identified, such as a particular basic phrase or set
of
information that is responsive to the received input. In traditional
conversational
interfaces, the initial response content would be provided to the user via the
conversational interface. In this solution, however, the responsive content,
may be
mapped to one or more synonym words or phrases, where the various synonym
words or
phrases are associated with predetermined scores for one or more measures. In
the
current example, the initial response content may be associated with a first
set of scores
for both formality and politeness. The synonym words or phrases may have
different sets
of scores for formality and politeness. Based on the determine scores
associated with the
received input, one or more synonym words or phrases having the same meaning
as
portions, or tokens, of the initial response content can replace those
corresponding
portions based on the synonym words or phrases having a relatively similar or
matching
scores as compared to the content of the received input. Once at least one
synonym is
used to replace at least a portion of the initial response content, the
modified and
personalized response can be output to the user through the conversational
interface.
[0035] In a second solution, the received input and particular information
about a
user are considered to apply persona-specific modifications to a set of
responsive content.
Well-known celebrities each convey particular different personas that can
affect the
behavior of a user, including by providing or generating an additional level
of trust for
interactions. An example of such a persona could be Chef Gordon Ramsey, who
may be
associated with strict instructional feedback that is intended to cause a
responsive action
quickly. Another example of a different type of persona may be Oprah Winfrey,
whose
9
CA 2981261 2017-10-04
persona may be used in a more motherly manner to provide encouragement or
kindly
words to a user, particularly where a note of worry is present in the received
input.
[0036] Similar to the first solution, the received input can be analyzed to
identify
a particular input personality type. In some instances, the input personality
type can be
determined based on a combination of various measurements determined from the
analyzed input, including formality, politeness, regional terms or dialects,
and other
measurements. When the received input is associated with a voice input, other
auditory
factors can be considered, including a pitch of the voice, a length of the
sounds, a
loudness of the voice, a timber of the voice, as well as any other suitable
factors. If the
input is associated with video input, additional facial recognition features
can be used to
provide additional context to the input. In response to these measurements,
the
conversational interface can identify a particular input personality type
associated with
the received input.
[0037] The particular input personality type can then be mapped or connected
to a
persona response type (e.g., intellectual, philosophical, motherly, strong,
etc.). Each
persona may also be associated with at least one persona response type, such
that a
determination that a particular persona response type should be applied to an
output may
mean that two or more personas may be available. The present solution,
however, can
identify one or more social network-based actions, preferences, or other
information
about one or more social network accounts associated with the user interacting
with the
conversational interface. Using this social network information, the
particular user's
likes, interactions, and follows, among other relevant social media content
and
interactions, can be used to determine one or more available celebrities or
entities with
which the user has previously interacted. Based on this information, which can
be stored
in a user profile associated with the user or otherwise accessed in response
to the
conversational interface interaction, a particular persona available within
the
conversational interface and corresponding to the persona response type can be
identified
and used to modify the response. Any suitable modification can be used to
modify and/or
enhance the response content, such as including and/or substituting one or
more common
phrases and/or words associated with the persona in place of the initial
response content.
CA 2981261 2017-10-04
In some instances, the output may be presented to the user in the voice
associated
with the particular persona. In other instances, an image or video associated
with the
persona may be included with the response content.
[0038] Turning to the illustrated example implementation, FIG. 1 is a block
diagram
illustrating an example system 100 for personalizing interactions with a
conversational
interface. System 100 includes functionality and structure associated with
receiving inputs
from a client device 180 (associated with a user), analyzing the received
input at the
conversational analysis system 102 to identify an intent of the input and
information on the
personality or internal context of the input, identifying a response
associated with the
received input, and identifying at least one manner in which the identified
response can be
personalized before transmitting the response back to the client device 180.
Specifically,
the illustrated system 100 includes or is communicably coupled with a client
device 180, a
conversational analysis system 102, one or more social networks 195, one or
more external
data sources 198, and network 170. System 100 is a single example of a
possible
implementation, with alternatives, additions, and modifications possible for
performing
some or all of the described operations and functionality. Although shown
separately, in
some implementations, functionality of two or more systems, servers, or
illustrated
components may be provided by a single system or server. In some
implementations, the
functionality of one illustrated system or server may be provided by multiple
systems,
servers, or computing devices, including those physically or logically local
or remote to
each other. Any combination or permutation of systems may perform the
functionality
described herein. In some instances, particular operations and functionality
described
herein may be executed at either the client device 180, the conversational
analysis system
102, or at one or more other non-illustrated components, as well as at a
combination
thereof.
[0039] As used in the present disclosure, the term "computer" is intended to
encompass any suitable processing device. For example, client device 180 and
the
conversational analysis system 102 may be any computer or processing device
(or
combination of devices) such as, for example, a blade server, general-purpose
personal
computer (PC), Mac , workstation, UNIXTm-based workstation, embedded system or
any
11
Date Recue/Date Received 2020-09-20
other suitable device. Moreover, although FIG. 1 illustrates particular
components as a
single element, those components may be implemented using a single system or
more than
those illustrated, as well as computers other than servers, including a server
pool or
variations that include distributed computing. In other words, the present
disclosure
contemplates computers other than general-purpose computers, as well as
computers
without conventional operating systems. Client device 180 may be any system
which can
request data, execute an application (e.g., client application 184), and/or
interact with the
conversational analysis system 102 and the conversational interface 108. The
client device
180, in some instances, may be any other suitable device, including a mobile
device, such
as a smartphone, a tablet computing device, a smartwatch, a laptop/notebook
computer, a
connected device, or any other suitable device. Additionally, the client
device 180 may be
a desktop or workstation, server, or any other suitable device. Similarly, the
conversational
analysis system 102 may be a server, a set of servers, a cloud-based
application or system,
or any other suitable system. In some instances, the client device 180 may
execute on or
be associated with a system executing the conversational analysis system 102.
In general,
each illustrated component may be adapted to execute any suitable operating
system,
including Linux, UNIXTM, WindowsTM, Mac OS , JavaTM, AndroidTM, Windows
PhoneTM
OS, or iOSTM, among others.
[0040] The conversational analysis system 102 can perform functionality
associated with one or more backend conversational interfaces 108, and can
perform
operations associated with receiving input from a client device 180 (e.g., via
conversational
interface 185) associated with the backend conversational interface 108, and
can analyze
the received input to determine an intent of the input (e.g., a particular
question, query,
comment, or other communication to which a response may be generated for the
conversational interface 108) and a personality associated with the input.
Based on the
determined intent, the conversational analysis system 102 can generate a
corresponding
response or response content to be provided back to the client device 180.
Using a natural
language generation engine 124, the conversational analysis system 102 can
generate an
appropriate initial response. Based on the personality associated with the
input, the
conversational analysis system 102 can identify or determine at least one
12
Date Recue/Date Received 2020-09-20
personalization to be performed on the initial response to be provided to the
client device
180, where applying the at least one personalization causes the initial
response to be
transformed into a personalized response transmitted to the client device 180
in response
to the received input.
[0041] As described, two different types of personalizations are possible with
the
described system. In a first solution, the personality of the input can be
assigned one or
more scores associated with various measures of the input, such as politeness
and
formality, among others. Based on the determined score, the response content
identified
by the system 102 can be modified to include or be associated with similar or
related
characteristics as those of the received input. As an example, a relatively
informal input
but with a relatively high level of politeness can cause the system to modify
at least some
of the response content with synonyms having matching levels of informality
and
politeness. In this way, the lexical personality of the response matches that
of the
received input.
[0042] In the second solution, the intent and personality of the received
input is
determined, where the input is associated with a particular user profile. A
set of
responsive content associated with the determined intent of the input is
identified, as well
as a persona response type that is based on the personality or state
identified from the
input. A set of social network activity information associated with the
particular user
profile can be identified, where the social network activity information
includes persons,
entities, and things which the particular user profile likes, follows, or has
interacted with
within at least one of the social networks. Based on that information, a
particular persona
that matches or corresponds to the persona response type and the social
network activity
information can be identified from a persona library. The particular persona
can be
associated with at least one persona-related content, where that content is
used to modify
at least a portion of the set of responsive content. Once the persona-specific
response is
generated it can be transmitted back to the client device 180 responsive to
the received
input.
[0043] As illustrated, the conversational analysis system 102 includes an
interface
104, a processor 106, a backend conversational interface 108, a natural
language
13
CA 2981261 2017-10-04
processing (NLP) engine 110, a response analysis engine 120, a natural
language
generation (NLG) engine 124, and memory 134. Different implementations may
include
additional or alterative components, with FIG. 1 meant to be an example
illustration of
one possible implementation. While illustrated separate from one another, at
least some
of these components, in particular the backend conversational interface 108,
the NLP
engine 110, the response analysis engine 120, and the NLG engine 124 may be
combined
within a single component or system, or may be implemented separate from one
another,
including at different systems and/or at remote components.
[0044] Interface 104 is used by the conversational analysis system 102 for
communicating with other systems in a distributed environment ¨ including
within the
environment 100 ¨ connected to the conversational analysis system 102 and/or
network
170, e.g., client device 180, social network 195, and/or any other external
data sources
198, as well as other systems or components communicably coupled to the
network 170.
Generally, the interface 104 comprises logic encoded in software and/or
hardware in a
suitable combination and operable to communicate with the network 170 and
other
communicably coupled components. More specifically, the interface 104 may
comprise
software supporting one or more communication protocols associated with
communications such that the conversational analysis system 102, network 170,
and/or
interface's hardware is operable to communicate physical signals within and
outside of
the illustrated environment 100.
[0045] Network 170 facilitates wireless or wireline communications between the
components of the environment 100 (e.g., between combinations of the
conversational
analysis system 102, client device(s) 180, and/or the other components, among
others) as
well as with any other local or remote computer, such as additional mobile
devices,
clients, servers, remotely executed or located portions of a particular
component, or other
devices communicably coupled to network 170, including those not illustrated
in FIG. 1.
In the illustrated environment, the network 170 is depicted as a single
network, but may
be comprised of more than one network without departing from the scope of this
disclosure, so long as at least a portion of the network 170 may facilitate
communications
between senders and recipients. In some instances, one or more of the
illustrated
14
CA 2981261 2017-10-04
components (e.g., the conversational analysis system 102) or portions thereof
(e.g., the
NLP engine 110, the response analysis engine 120, the NLG engine 124, or other
portions)
may be included within network 170 as one or more cloud-based services or
operations.
The network 170 may be all or a portion of an enterprise or secured network,
while in
another instance, at least a portion of the network 170 may represent a
connection to the
Internet. In some instances, a portion of the network 170 may be a virtual
private network
(VPN) or an Intranet. Further, all or a portion of the network 170 can
comprise either a
wireline or wireless link. Example wireless links may include
802.11a/b/g/n/ac, 802.20,
WiMaxTm, LTE, and/or any other appropriate wireless link. In other words, the
network
170 encompasses any internal or external network, networks, sub-network, or
combination
thereof operable to facilitate communications between various computing
components
inside and outside the illustrated environment 100. The network 170 may
communicate,
for example, Internet Protocol (IP) packets, Frame Relay frames, Asynchronous
Transfer
Mode (ATM) cells, voice, video, data, and other suitable information between
network
addresses. The network 170 may also include one or more local area networks
(LANs),
radio access networks (RANs), metropolitan area networks (MANs), wide area
networks
(WANs), all or a portion of the Internet, and/or any other communication
system or systems
at one or more locations.
[0046] The conversational analysis system 102 also includes one or more
processors 106. Although illustrated as a single processor 106 in FIG. 1,
multiple
processors may be used according to particular needs, desires, or particular
implementations of the environment 100. Each processor 106 may be a central
processing
unit (CPU), an application specific integrated circuit (ASIC), a field-
programmable gate
array (FPGA), or another suitable component. Generally, the processor 106
executes
instructions and manipulates data to perform the operations of the
conversational analysis
system 102, in particular those related to executing the various modules
illustrated therein
and their related functionality. Specifically, the processor 106 executes the
algorithms and
operations described in the illustrated figures, as well as the various
software modules and
functionalities, including the functionality for sending communications to and
receiving
transmissions from the client device 180 and the social
Date Recue/Date Received 2020-09-20
network(s) 195, as well as to process and prepare responses to received input
associated
with the conversational interface 108. Each processor 106 may have a single
core or
multiple cores, with each core available to host and execute an individual
processing
thread.
[0047] Regardless of the particular implementation, "software" includes
computer-
readable instructions, firmware, wired and/or programmed hardware, or any
combination
thereof on a tangible medium (transitory or non-transitory, as appropriate)
operable when
executed to perform at least the processes and operations described herein. In
fact, each
software component may be fully or partially written or described in any
appropriate
computer language including C, C++, Objective-C, JavaScript, JavaTM, Visual
Basic,
assembler, Peri , Swift, HTML5, any suitable version of 4GL, as well as
others.
[0048] As illustrated, the conversational analysis system 102 includes, is
associated
with, and/or executes the backend conversational interface 108. The
backend
conversational interface 108 may be a program, module, component, agent, or
any other
software component which manages and conducts conversations and interactions
via
auditory or textual methods, and which may be used to simulate how a human
would
behave as a conversational partner. In some instances, the backend
conversational interface
108 may be executed remotely from the conversational analysis system 102,
where the
analysis system 102 performs operations associated with determining and
modifying the
personality of certain inputs and responses, but where the conversational
interface 108
determines the intent of the response and/or the responses to be provided. The
backend
conversational interface 108 may be accessed via a website, a web service
interaction, a
particular application (e.g., client application 184), or it may be a backend
portion of a
digital or virtual assistant application or functionality of a particular
operating system, such
as AppleTM Inc.'s Siri TM, GoogleTM LLC's Google Assistant, AmazonTM
Technologies
Inc.'s Alexa, MicrosoftTM Corporation's CortanaTM, or others. In some
instances, a remote
agent or client-side portion of the conversational interface 185 may execute
at the client
device 180, where inputs can be provided and responses presented to the user
of the client
device 180, while some or all
16
Date Recue/Date Received 2020-09-20
of the processing is performed at the conversational analysis system 102 and
the backend
conversational interface 108.
[0049] The NLP engine 110 represents any suitable natural language processing
engine, and performs operations related to understanding a set of received
input received
at the backend conversational interface 108. Examples of NLP engines that
could be used
or implemented include a plurality of web services and backend applications,
including
WatsonTM by IBMTm, GoogleTM LLC's Cloud Natural Language API, Amazon Lex ,
Microsoft Cognitive Services, as well as any proprietary solution,
application, or service.
The processing performed by the NLP engine 110 can include identifying an
intent
associated with the input received via the conversational interface 108, which
is performed
by the intent deciphering module 112. The result of the intent determination
can be a set
of lexical semantics of the received input, which can then be provided to the
response
analysis engine 120, where at least an initial set of response content can be
generated.
[0050] As illustrated, the NLP engine 110 further includes a personality
deciphering
module 114, where the personality deciphering module 114 can perform one or
more
operations to identify a particular personality of the received input. Any
suitable
algorithms can be used to determine the personality of the input. In some
instances, scores
for multiple measures of the input can be generated, including scores related
to a formality
of the input, a politeness of the input, usage of particular regional phrases,
an accent or
particular phrasing associated with the input, a level of sarcasm within the
input, an
emotional state associated with the input, as well as any other suitable
measures or values.
The personality deciphering module 114 may use, in particular, a lexical
personality
scoring engine 118 to generate the scores based on one or more rule sets and
analysis
factors. In some instances, an analysis engine 116 may be used to analyze the
type of input
received. The analysis engine 116 may be able to analyze and assist in the
evaluation and
scoring of textual input, auditory input, video or still image input (e.g.,
associated with a
facial analysis or with a read of the eyes). For example, textual input may be
analyzed to
identify a score or evaluation of a textual tone or syntax, the politeness and
formality of
the syntax, a level of sarcasm, and particular vocabulary used. An auditory
input, which
can include the voice of a user associated with the received
17
Date Recue/Date Received 2020-09-20
input, can be analyzed to further obtain information on a pitch of the voice,
a length of
sounds, a loudness of the voice, a timber of the voice, any shakiness or
quavering of the
voice, as well as other measures. If a video or image of the user associated
with the input
is available, such as from a camera associated with the client device 180,
facial analysis
techniques and body language-related analyses can be used to determine
particular stress,
upsetness, and other emotionally-related measures. Based on these analyses,
one or more
scores or evaluations associated with the personality of the input can be
obtained or
generated.
[0051] In some instances, based on a combination of the scores, a personality
input type can be identified or determined. The personality input types can be
associated
with predetermined values or ranges of values for different measure
combinations.
Example personality input types may include "scared," "calm, inquisitive,"
"calm,
instructional," "angry," and "angry, unsure," among others. In some instances,
the lexical
personality scoring engine 118 can, in addition to analyzing and determining
scores or
evaluations of the different available measures, can identify and assign a
particular
personality input type to the current received input. In some instances, the
lexical
personality scoring engine 118, or another component, can also map or identify
a
particular persona response type to the personality input type, where the
persona response
types identify a type of persona to be associated with and to be used in
modifying a set of
response content as described herein. The persona response types can include
any
suitable categories, including, for example, "intellectual," "philosophical,"
"motherly,"
and "strong," among others. In some instances, the persona response type may
also be
determined, at least in part, on a particular action that the conversational
interface 108
and the response analysis engine 120 determine that the user should be
recommended
into performing. For example, in a financial-related conversational interface
108, a
question about a particular transaction for a user may be determined as the
intent of the
received input, in response to processing the response in light of a
particular user's
accounts or financial situation, a determination may be made by the response
analysis
engine 120 to encourage the user not to continue with the transaction. In
those instances,
the particular persona response type used to enhance the response may be a
harsher or
18
CA 2981261 2017-10-04
stricter persona response type. For instance, instead of choosing a motherly
persona
response type, a stricter or stronger persona response type can be used to
emphasize a
recommendation not to complete the transaction.
[0052] As mentioned, the determined intent of the received input can be
provided
to a response analysis engine 120, where the response analysis engine 120
determines a
particular ideal response or response sentiment that should be used as a basis
for the
response. In some instances, where the determined intent of the received input
is a
question, the response analysis engine 120 may be used to derive a responsive
answer to
the question. The answer may be a particular value or query response, and can
be provided
to the NLG engine 124 for further processing and generation of a sentence or
other detailed
response to be provided via the conversational interface 108. The response
analysis engine
120 may be any suitable component or system capable of generating a response
to the
determined intent of the received intent. In some instances, the response
analysis engine
120 may be associated with a search engine, an expert system, a knowledge
base, or any
other suitable backend component able to respond to the input.
[0053] In some instances, the response analysis engine 120 can include a user
profile analysis engine 122, where additional information about a particular
user profile
associated with the received input can be obtained. The user profile analysis
engine 122
can be used to provide a user-specific response, and can use information
associated with a
particular user profile within a plurality of user profiles 158. In some
instances, the user
profile 158 can identify a set of preferences 160 previously defined by the
user or
determined based on previous interactions and other user operations. As
illustrated, social
network data 162 may be included or associated with the user profile 158. In
some
instances, the social network data 162 may identify particular social network
profiles
and/or login credentials associated with one or more social networks 195. The
user profile
analysis engine 122 can use that information to access a particular social
network 195 (e.g.,
FacebookTM, TwitterTm, RedditTM, InstagramTM, etc.) and obtain information
about
particular social network activities of the user, including particular
persons, entities, and
other accounts with which the user follows, likes, or otherwise has interacted
with as
19
Date Recue/Date Received 2020-09-20
defined by a social network user profile 196, where that information
identifies particular
public personalities or entities with which later persona-based content and
enhancements
can be used. In some instances, at least some of the information may be
available within
the social network data 162 without requiring the user profile analysis engine
122 or
another component to access the social networks 195 at the time of the
interaction with
the conversational interface 108, instead accessing the relevant social
network data.
Further, the user profile 158 may store financial data 164 associated with the
user profile
158, including transaction histories, current account information, and other
similar data.
In some instances, information to access such information may be stored in the
financial
data 164, and the user profile analysis engine 122 can use that data to access
the accounts
or other relevant data in real-time during the interactions with the interface
108. In some
instances, location information associated with the user may be included in
the user
profile 158, or may be accessible via the client device 180 or other means.
The location
information can be used to personalize responses, and to include that
information as part
of the response analysis.
[0054] The NLG engine 124 can receive the output of the response analysis
engine 120 and prepare a natural language response to the received input based
on that
output. The NLG engine 124 can be any suitable NLG engine capable of
generating
natural language responses from the output of the response analysis engine
120. In some
instances, the NLG engine 124 can identify or otherwise determine at least a
base set of
words, phrases, or other combinations or tokens to be used in representing the
response
content received from the response analysis engine 120. The base set of words,
phrases,
or tokens are associated with an initial response determined by the response
analysis
engine 120, and may be an initial representation of the natural language
result. The
syntax and semantic generation module 126 can perform these operations, and
can make
decisions on how to place the concept of the response received from the
response analysis
system 120 into a natural language set of words, phrases, or combination of
tokens.
[0055] Once the initial set of response content is available, the lexical
personality
filter module 128 can be used to determine and apply the appropriate
modifications to the
response content to be used to generate a personalized response. As noted, the
CA 2981261 2017-10-04
personalized response can be generated by identifying one or more synonyms
(e.g., from
the synonym repository 136) of the base words, phrases, or combinations of
tokens
included in the initial response and replacing one or more of those base
portions with
synonyms having a matching or similar set of characteristics as those
determined to be
associated with received input using a personality-based response module 130.
The
personality-based response module 130 can use a synonym repository 136 to
identify
entries associated with particular base words (or tokens) 138. Each of the
base words 138
may be associated with a plurality of synonyms 140, where each synonym 140 is
associated with a set of one or more predefined lexical personality scores
142. Those
scores 142 can be compared to the scores associated with the received input as
determined in the lexical personality scoring engine 118, and one or more of
the
synonyms 140 identified as appropriate for substitution based on their match.
The lexical
personality filter module 128 can then modify the initial response content by
replacing
the base words 138 with the at least one identified synonyms 140.
[0056] The lexical personality filter module 128 can also include and use a
persona module 132 to apply a particular persona to a set of response content.
The
persona module 132 can access a persona library 144 storing one or more
persona entities
146. The persona entities 146 can be associated with celebrities and other
well-known
persons, where each entity 146 is associated with one or more persona response
types
148. The persona response types 148 can be matched or associated with the
persona
response type determined by the personality deciphering module 114 to identify
which
persona entities 146 to associate with a particular response. Importantly, the
described
process can use the social network data 162 associated with a particular user
profile 158
to determine a subset of persona entities 146 to consider for application to a
particular
response based on those likes, follows, or other interactions performed by the
user. From
that subset of persona entities 146, one or more persona entities 146 that
match the
persona response type can be identified, and a best or preferred persona can
be
determined.
[0057] Each persona entity 146 can be associated with a set of persona phrases
and words 150 and/or a set of persona-related content 154. The set of persona
phrases
21
CA 2981261 2017-10-04
and words 150 may be well-known phrases or language used by the celebrity or
other
person in the real world, such that those phrases or language can be used in
responses. In
some instances, those phrases and words 150 can be associated with one or more
synonyms or base words 152, where the base words 152 can be used to identify
relevant
phrases and words 150 to be used. Modifying the initial set of response
content can
include replacing one or more base words or tokens with those well-known words
or
phrases. Alternatively, the modification can include supplementing or adding
some of
those well-known words or phrases alongside the initial response content,
sometimes
without substituting any of the initial phrases. In some instances, the
persona phrases and
words 150 may be associated with one or more lexical personality scores, such
that those
persona phrases and words 150 that correspond to the input personality score
can be used,
combining the two solutions.
[0058] The set of persona-related content 154 can include additional non-
phrase
or word content that can be used to personalize a response or conversational
interface 108
with the identified persona. In some instances, such content 154 can include
images,
video, or other media associated with the persona entity 146, where that
content 154 can
be presented along with a generated response. As illustrated, the content 154
may include
audio content 156 in the voice of the person associated with the persona
entity 146, such
that at least a portion of the response transmitted to the client device 180
can be provided
in the same voice of the persona. Other unique or defining aspects of the
particular
persona can be included in the persona-related content 154 and used to
embellish or
personalize the response.
[0059] As illustrated, the conversational analysis system 102 includes memory
134. In some implementations, the conversational analysis system 102 includes
a single
memory or multiple memories. The memory 134 may include any type of memory or
database module and may take the form of volatile and/or non-volatile memory
including, without limitation, magnetic media, optical media, random access
memory
(RAM), read-only memory (ROM), removable media, or any other suitable local or
remote memory component. The memory 134 may store various objects or data,
including caches, classes, frameworks, applications, backup data, business
objects, jobs,
22
CA 2981261 2017-10-04
web pages, web page templates, database tables, database queries, repositories
storing
business and/or dynamic information, and any other appropriate information
including
any parameters, variables, algorithms, instructions, rules, constraints, or
references
thereto associated with the purposes of the conversational analysis system
102.
Additionally, the memory 134 may store any other appropriate data, such as VPN
applications, firmware logs and policies, firewall policies, a security or
access log, print
or other reporting files, as well as others. As illustrated, memory 134
includes, for
example, the synonym repository 136, the persona library 144, and the user
profiles 158,
described herein. Memory 134 may be local to or remote to the conversational
analysis
system 102, and may be available remotely via network 170 or an alternative
connection
in such instances where not locally available. Further, some or all of the
data included in
memory 134 in FIG. 1 may be located outside of the conversational analysis
system 102,
including within network 170 as cloud-based storage and data.
[0060] Illustrated system 100 includes at least one client device 180, and may
include a plurality of client devices 180 in some instances. Each client
device 180 may
generally be any computing device operable to connect to or communicate within
the
system 100 via the network 170 using a wireline or wireless connection. In
general, the
client device 180 comprises an electronic computer device operable to receive,
transmit,
process, and store any appropriate data associated with the system 100 of FIG.
1. As
illustrated, the client device 180 can include one or more client
applications, including
the client application 184 and a digital assistant 186. In some instances, the
digital
assistant 186 may be a part of the operating system executing on the client
device 180, or
may be a standalone application or client-side agent of a backend application.
In some
instances, the client device 180 may comprise a device that includes one or
more input
devices 187, such as a keypad, touch screen, camera, or other device(s) that
can interact
with the client application 184 and/or digital assistant 186 and other
functionality, and
one or more output devices 188 that convey information associated with the
operation of
the applications and their application windows to the user of the client
device 180. The
output devices 188 can include a display, speakers, or any other suitable
output
component. The information presented by the output devices 188 can include
digital
23
CA 2981261 2017-10-04
data, visual information, auditory output, or a graphical user interface (GUI)
183, as
shown with respect to the client device 180. In general, client device 180
comprises an
electronic computer device operable to receive, transmit, process, and store
any
appropriate data associated with the environment 100 of FIG. 1.
[0061] Client application 184 can be any type of application that allows the
client
device 180 to request and view content on the client device 180. In some
instances, client
application 184 may correspond with one or more backend applications or
functionality,
including an application or platform associated with the conversational
analysis system
102. In some instances, the client application 184 can be associated with a
client-side
version of the conversational interface 185, where the client-side version of
the
conversational interface 185 can represent a means for users to provide inputs
to the
conversational interface 108 and receive the personalized output of the same
for viewing
at the client device 180.
[0062] In many instances, the client device 180 may be a mobile device,
including
but not limited to, a smartphone, a tablet computing device, a laptop/notebook
computer,
a smartwatch, or any other suitable device capable of interacting with the
conversational
analysis system 102 and the conversational interface 108. One or more
additional client
applications 184 may be present on a client device 180, and can provide
varying
functionality for users. In some instances, client application 184 may be a
web browser,
mobile application, cloud-based application, or dedicated remote application
or software
capable of interacting with at least some of the illustrated systems via
network 170 to
request information from and/or respond to one or more of those systems.
[0063] The digital assistant 186 may be any interactive artificial or virtual
intelligence component, agent, or other functionality that can be interacted
with by a user,
either textually or verbally through one or more input components 187 (e.g., a
microphone), manually through one or more input components 187 (e.g., physical
or
virtual keyboards, touch screen buttons or controls, other physical or virtual
buttons,
etc.), or through captured gestures or movements identified by the client
device 180. In
general, the digital assistant 186 may be a software agent, module, or
component, among
others, that can perform tasks or services for an individual in response to
one or more
24
CA 2981261 2017-10-04
inputs, and can include or represent a particular conversational interface
associated with
the backend conversational interface 108. As indicated, any one of numerous
commercial
examples may be used, as well as other proprietary or application-specific
assistants. The
digital assistant 186 may work and interact via text (e.g., chat), voice,
image submission,
or other suitable inputs. Some virtual assistants can interpret input using
natural language
processing (NLP) to match user text or voice input to executable commands. In
some
instances, the digital assistant 186 can be interacted with to initiate and
perform one or
more input and response interactions described herein. In some instances, the
digital
assistant 186 may be a standalone application (e.g., GoogleTM LLC's Assistant
executing
on an iPhoneTm), functionality included in a particular application used for
other purposes
(e.g., an Alexa-enabled AmazonTM app), or an agent or other functionality
built into the
operating system (e.g., SiriTM on AppleTm Inc.'s iOSTm).
[0064] As illustrated, the client device 180 may also include an interface 181
for
communication (similar to or different from interface 104), a processor 182
(similar to or
different from processor 106), memory 189 (similar to or different from memory
134), and
GUI 183. GUI 183 can interface with at least a portion of the environment 100
for any
suitable purpose, including generating a visual representation of the client
application 184
and/or the digital assistant 186, presenting a pop-up or push notification or
preview thereof,
presenting the UI associated with the conversational interface 185, or any
other suitable
presentation of information. GUI 183 may also be used to view and interact
with various
Web pages, applications, and Web services located local or external to the
client device
180, as well as information relevant to the client application 184. Generally,
the GUI 183
provides the user with an efficient and user-friendly presentation of data
provided by or
communicated within the system. The GUI 183 may comprise a plurality of
customizable
frames or views having interactive fields, pull-down lists, and buttons
operated by the user.
For example, the GUI 183 may provide interactive elements that allow a user to
view or
interact with information related to the operations of processes associated
with the
conversational analysis system 102 and any associated systems, among others.
In general,
the GUI 183 is often configurable, supports a combination of tables and graphs
(bar, line,
pie, status dials, etc.), and is able to build real-time portals,
Date Recue/Date Received 2020-09-20
application windows, and presentations. Therefore, the GUI 183 contemplates
any
suitable graphical user interface, such as a combination of a generic web
browser, a web-
enabled application, intelligent engine, and command line interface (CLI) that
processes
information in the platform and efficiently presents the results to the user
visually.
[0065] The external data sources 198 illustrated in FIG. 1 may be any other
data
source that provides additional information to the conversational analysis
system 102.
The information may be used by the response analysis engine 120 to determine a
particular response to the received input, or the information may be used to
personalize
the response as described herein. Any number of data sources 198 may be used
in
alternative implementations.
[0066] While portions of the elements illustrated in FIG. 1 are shown as
individual modules that implement the various features and functionality
through various
objects, methods, or other processes, the software may instead include a
number of sub-
modules, third-party services, components, libraries, and such, as
appropriate.
Conversely, the features and functionality of various components can be
combined into
single components as appropriate.
[0067] FIGS. 2A and B combine to represent an illustration of a data and
control
flow of example components and interactions 200 performed by a system
performing
personalization operations with a conversational interface related to an
analysis of the
lexical personality of an input, where the responsive output is provided with
an output
lexical personality based on the input lexical personality. The diagram
provides an
example set of operations and interactions to identify the personality with
the received
input and to prepare and provide a personalized response with the same or
similar
personality. Illustrated in FIGS. 2A-2B are a user 201 interacting with a
conversational
interface (or client device at which a user is interacting with the
conversational interface),
an NLP engine 208 (including an intent deciphering module 210 and a
personality
deciphering module 215), a response analysis engine 225, and an NLG engine
228.
These components may be similar to or different the components described in
FIG. 1.
[0068] As illustrated, the user 201 provides a conversational or user input
205 to
the NLP engine 208. The NLP engine 208 determines both an intent of the
26
CA 2981261 2017-10-04
conversational input 205 (using the intent deciphering module 210) and a
determination
of a lexical personality score or rating associated with the content of the
conversational
input 205 (using the personality deciphering module 215). In some instances,
the
determination of the intent and the determination of the lexical personality
score may be
separate activities or actions. In other instances, however, the determined
intent may
impact or otherwise affect the determination of the lexical personality score,
while the
lexical personality score may also impact a determination of intent. The
intent can
represent the question, query, or information associated with the
conversational input
205, and can be determined using any suitable method. In some instances, a
natural
language understanding (NLU) engine may be used to understand the intent,
which is
what the conversational input 205 wants, is asking about, or otherwise relates
to. The
intent of the conversational input 205 is then used to determine a suitable
response.
[0069] The personality deciphering module 215 can evaluate the particular
conversational input 205 using one or more measures, where each measure is
graded on a
suitable scale. As illustrated in input table 220 (expanding in FIG. 2B),
example
measures can include a level of formality and a level of politeness associated
with the
conversational input 205. The personality deciphering module 215 can generate
scores
associated with the analysis. As illustrated, the input table 220 shows a
plurality of
potential conversational inputs of differing levels of formality and
politeness. In this
example, the conversational input 205 can be represented by sentence 221,
which recites
"My bad, I can't find it, please help me out and I'll owe you one." Based on
analysis of
the vocabulary and syntax used within the sentence, a formality score of 0.1
(on a scale
from 0 to 1) and a politeness score of 0.95 (on a scale of 0 to 1) are
generated. Input
table 220 illustrates formality and politeness scores associated with other
sentences
having the same intent, and shows how various inputs can be scored. Based on
the
scoring analysis, the determined lexical personality score(s) are associated
with the
conversational input 205.
[0070] The NLP engine 208 provides its outputs of the determined intent and
the
lexical personality score(s) to the response analysis engine 225. The engine
225 uses this
information to determine a suitable response to the conversational input 205
based on the
27
CA 2981261 2017-10-04
intent. In some instances, the lexical personality score(s) may be provided to
the lexical
personality filter module 240 for later use, as the engine 225 may or may not
use or apply
the score(s) to determine the suitable response.
[0071] Once the essence or general substance of the response is determined,
the
response analysis engine 225 provides that information to the NLG engine 228,
where the
NLG engine 228 is used to prepare the conversational response based on the
response
essence or content identified by the response analysis engine 225. As
illustrated, the
NLG engine 228 includes a syntax and semantic generation module 230 and the
lexical
personality filter module 240. The syntax and semantic generation module 230
can be
used to generate at least a base set of tokens (236 as illustrated in FIG. 2B)
as an initial
natural language response to the conversational input 205. The tokens may each
represent a single word or phrase associated with an initial response. For
example,
several tokens may be generated in response to the example input above, and
may
indicate that based on the determined intent, the set of tokens may include "I
will" "help"
"you find it, [Sir or Miss]."
[0072] Once the initial response is determined, the lexical personality filter
module 240 can be used to determine at least one personalization to be made to
the initial
response tokens based on the prior determination of the lexical personality
scores. The
lexical personality filter module 240, for instance, can identify a set of
synonyms for at
least some of the tokens included in the base words 236. In some instances,
synonyms
may be identified for particular tokens, while in other instances, synonyms
may be
identified for the entire sentence or response. As illustrated in synonym
table 235, each,
or at least some, of the base token or word can be associated with one or more
synonyms
or synonym tokens. Each synonym or synonym token may be associated with a
corresponding lexical personality score. As illustrated in synonym table 235,
a similar
scoring methodology can be used for the synonyms as the conversational input
205 in the
input table 220. Based on the determined scores for the conversational input
205, one or
more synonym tokens can be identified as having similar lexical personality
scores. In
the illustrated example, the second set of synonyms 237 are associated with a
formality
score of 0.1 and a politeness score of 0.9, relatively similar to the scores
of the
28
CA 2981261 2017-10-04
conversational input 205 determined at the NLP engine 208 (i.e., 0.1 for
formality and
0.95 for politeness). Therefore, that set of synonyms 237 can be identified as
the
appropriate modification for a personalized response. In some instances, each
individual
synonym token may be associated with lexical personality scores, such that the
identification of synonyms is done on a base token by base token basis. In
some
instances, such as where several synonym tokens are associated with similar
lexical
personality scores, two or more synonym tokens associated with different
scores may be
used in personalizing a particular response. For example, if the politeness
score of the
conversational input 205 was 0.8, some of the synonyms in the second set of
synonyms
237 may be used as well as some of the synonyms in the third set of synonyms
238.
[0073] Once the suitable synonyms matching the personality of the
conversational
input 205 are identified, the NLG engine 228 can replace at least one token
from the
initial token set with the at least one suitable synonym. In some instances,
only one of
the initial tokens may be replaced or substituted with a synonym token, while
in others
multiple tokens or the entire set of tokens may be replaced. The modified
response
represents a conversational response 245 from the NLG engine 228, which can
then be
transmitted back to the user 201 (or the client device associated with user
201).
[0074] FIG. 3 is a flow chart of an example method 300 performed at a system
associated with a conversational interface to identify a first lexical
personality of an input
and provide a response to the input by applying a second lexical personality
based on the
identified first lexical personality. It will be understood that method 300
and related
methods may be performed, for example, by any suitable system, environment,
software,
and hardware, or a combination of systems, environments, software, and
hardware, as
appropriate. For example, a system comprising a communications module, at
least one
memory storing instructions and other required data, and at least one hardware
processor
interoperably coupled to the at least one memory and the communications module
can be
used to execute method 300. In some implementations, the method 300 and
related
methods are executed by one or more components of the system 100 described
above
with respect to FIG. 1, or the components described in FIGS. 2A-2B.
29
CA 2981261 2017-10-04
[0075] At 305, a first signal including a conversational input associated with
interactions with a conversational interface are received via a communications
module.
In some instances, the first signal can include only the content of the
conversational
input, while in other instances the first signal may also include an
identification of a
particular user or user profile associated with the conversational input. In
some
instances, the conversational input may be received at a specific endpoint
associated with
a particular conversational interface or related application, such that the
analysis of the
conversational input is based on the use of the particular conversational
interface or
related application. Examples may include a digital assistant or a particular
website,
where the conversational input is to be responded to based on the information
obtainable
by or related to that particular interface or application.
[0076] In some instances, the conversational input may be or may include text-
based input, auditory input (e.g., an audio file), and/or video or image
input. The
conversational input can be submitted by a user or agent associated with a
particular user
profile, although for purposes of some implementations of method 300,
identification of
the particular user profile is not needed to perform the personalization
process. In some
instances, the received conversational input can represent a particular query,
question, or
interactive communication.
[0077] At 310, the received conversational input is analyzed via a natural
language processing engine or service to determine an intent associated with
the input
and a lexical personality score based on the content of the received input.
The intent
associated with the input can be determined based on any suitable factor, and
can be used
to determine a meaning of a particular request associated with the
conversational input.
The intent may be used to determine a particular response to be generated. The
lexical
personality score can be used to identify a personality of the conversational
input, and to
modify the corresponding conversational response in a personalized manner such
that the
lexical personality of the response corresponds to the lexical personality of
the
conversational input. Determining the lexical personality score can be based
on any
suitable analysis techniques, and can result in scores being applied to the
conversational
input based on one or more measures or factors. In one example, scores may be
defined
CA 2981261 2017-10-04
for formality and politeness for a particular input. Scores may also be
determined for
sarcasm detected in the input, particular predefined words such as curse
words, words
associated with particular moods (e.g., "angry", "mad"), particular regional
words or
phrases, as well as other scores. Where the conversational input is a spoken
or verbal
input, the voice of the speaker can also be analyzed to identify a particular
accent, a mood
detectable from the speech patterns, and other audible-related aspects. Scores
for some
or all of these measures may be associated with the conversational input.
[0078] At 315, a determination is made as to a set of response content
responsive
to the determined intent of the received conversational input. In some
instances, the set
of response content can include a set of initial tokens to be used in an
initial or proposed
response. In some instances, the set of initial tokens can represent an
initial sentence or
phrase responsive to the conversational input, where the initial sentence or
phrase is
associated with a default lexical personality (e.g., where formality and
politeness are
scored at 1 on a 0 to 1 value range).
[0079] Once the initial tokens are identified, at 320 a set of synonym tokens
associated with at least some of the set of initial tokens can be identified.
In some
instances, a synonym repository may be accessed to identify synonyms
associated with at
least some of the initial tokens. The synonym tokens, which may represent
single words
or phrases corresponding to the initial token to which they are related, can
be associated
with at least one predetermined lexical personality score. In some instances,
the
predetermined lexical personality scores may be associated with measures that
are
evaluated and determined at 310. In other instances, at least some of the
predetermined
lexical personality scores may differ from the measures evaluated at 310.
[0080] At 325, at least one synonym token from the identified set of synonym
tokens can be determined to be associated with a lexical personality score
similar to the
determined lexical personality score of the received conversational interface.
In some
instances, the scores may exactly match, while in others, the scores may be
within a
particular threshold or range near or within the other scores. Any suitable
algorithm or
comparison operation can be used to determine which synonym tokens are close
enough
to a match to the lexical personality scores of the initial tokens.
31
CA 2981261 2017-10-04
[0081] At 330, at least one token from the set of initial tokens can be
replaced
with at least one determined synonym token (at 325). By replacing the initial
token with
the synonym token, a personalized version of the set of response content is
generated,
such that a personalized conversational response can be provided back to the
user.
[0082] At 335, a second signal including the personalized version of the set
of
response content is transmitted, via the communications module, to the
conversational
interface associated with the received first signal. In some instances, the
personalized
response may be modified to match or closely align with the lexical
personality score of
the conversational input initially received. In other instances, the lexical
personality of
the personalized response may differ from the lexical personality score of the
received
conversational input in a way that complements or otherwise corresponds to the
personality of the received input.
[0083] FIG. 4 is an illustration of a data and control flow of example
interactions
400 performed by a system performing persona-based personalization operations
for a
conversational interface based on an analysis of the lexical personality of an
input and a
corresponding user profile, generating a responsive output associated with the
input, and
modifying or transforming the responsive output to a persona-specific
responsive output
based on the corresponding user profile associated with the received input.
Illustrated in
FIG. 4 are a user 401 interacting with a conversational interface (or client
device at which
a user is interacting with the conversational interface), an NLP engine 408
(including an
intent deciphering module 410 and a personality deciphering module 415), a
response
analysis engine 425, and an NLG engine 428. These components may be similar to
or
different from the components described in FIG. 1, as well as those
illustrated in FIGS.
2A-2B.
[0084] As illustrated, the user 401 provides a conversational or user input
405 to
the NLP engine 408. As previously described, the NLP engine 408 can determine
an
intent of the conversational input 405 (e.g., using the intent deciphering
module 410). In
this solution, the NLP engine 408 can also identify one or more lexical
personality scores
or ratings associated with the content of the conversational input 405 (using
the
personality deciphering module 415). Alternatively, or based on those scores,
each
32
CA 2981261 2017-10-04
received conversational input 405 can be evaluated to determine a particular
personality
input type. The personality input type can describe a mood, tone, or other
aspect of the
conversational input and the user 401. In some instances, a combination of the
lexical
personality scores can be used to identify the personality input type, while
in others, the
personality input type may be determined directly without the scores. In
combination
with the personality input type, a corresponding persona response type can be
identified.
As illustrated in the personality input type and persona response type table
420, tables of
lexical score combinations can be considered by the personality deciphering
module 415
to determine the personality input type and the corresponding persona response
type. In
some instances, the persona response type may be mapped to one or more
different
personality input types, which in some instances can be stored separate from
the
personality input type table 420 and the corresponding lexical personality
scores. In the
illustrated example, lexical personality scores are determined for a
politeness level, a
pitch of a voice, a length of the sounds used in the input, a loudness of the
voice, and a
timber of the voice used in the conversational input 405. Different scores, or
a
combination of different scores, can result in the association of the
conversational input
with different personality input types. The illustrated rows of the table 420
can represent
only a portion of the possible score combinations and personality input types.
In the
illustrated example, a determination is made that the conversational input 405
corresponds to a "Scared" input personality type, such that a "Motherly"
persona
response type should be used, if available, to personalize the conversational
response.
[0085] The NLP engine 408 provides its outputs of the determined intent and
the
lexical personality score(s) and persona response type determination to the
response
analysis engine 425. The response analysis engine 425 uses this information to
determine
a suitable response to the conversational input 405 based on the intent. In
some
instances, the lexical personality score(s) and the persona response type
determination
may be provided to the lexical personality filter module 435 for later use, as
the response
analysis engine 425 may or may not use or apply the score(s) and persona
response type
determination to determine the suitable response.
33
CA 2981261 2017-10-04
[0086] As illustrated, the response analysis engine 425 can further identify
at least
some additional response-relevant information associated with the user or user
profile
associated with the interactions with the conversational interface. As
illustrated, the
response analysis engine 425 can interact with a user profile analysis engine
427 to
determine personalizations to be applied or additional contextual information
that can be
used in determining the appropriate response. The user profile analysis engine
427 can
include various sets of data or information about the user, as well as
information on how
to access remotely located information associated with the user.
[0087] The user profile analysis engine 427 can access or obtain social
network
data 450 associated with the user profile. The information can include
specific
information about one or more entity, business, or celebrity accounts with
which the user
follows, likes, or "hearts", among other indications, on one or more social
networks. In
some instances, the information can also include, or can be used to derive,
relatively
positive or negative interactions or mentions associated with those accounts
outside of
explicit indications of liking or following, such as positive status updates
or comments
associated with those entities or persons. Based on this information, the user
profile
analysis engine 427 can identify one or more accounts associated with the user
profile,
which may be used or associated with at least one persona available in a
persona library
associated with the lexical personality filter module 435. This information
can be shared
with the lexical personality filter module 435 and used to personalize the
output for the
user.
[0088] The user profile analysis engine 427 can also obtain or determine
contextual data 452 associated with the user, such as a current location or
other similar
information. The contextual data 452 can be considered in localizing or
contextualizing
the response content to the current status of the user. The user profile
analysis engine 427
can also consider one or more defined or derived preferences 456 of the user
profile. The
preferences 456 may be used to affect or impact the response substance
determined by
the response analysis engine 425, as well as to determine or identify a
particular persona
to be applied to a response by the lexical personality filter module 435.
34
CA 2981261 2017-10-04
[0089] The user profile analysis engine 427 can identify a financial status
454
associated with the user profile, where the financial status 454 can be used
to affect the
response to be provided back to the user. For example, if the conversational
input 405
relates to a financial transaction or a purchase, the financial status 454 can
be used to
update the response to be provided, such as by evaluating a recommendation or
particular
action to be taken. The result of the analysis in light of the financial
status 454 can be
used to determine the proper substance of the response.
[0090] Once the essence or substance of the response is determined, as well as
any contextual data or user profile-related data, the response analysis engine
425 can
provide that information to the NLG engine 428, where the NLG engine 428 is
used to
prepare the conversational response based on the response essence or content
identified
by the response analysis engine 425 and the persona response type identified
by the NLP
engine 408. As illustrated, the NLG engine 428 includes a syntax and semantic
generation module 430 and the lexical personality filter module 435. The
syntax and
semantic generation module 430 can be used to generate an initial response to
be
provided in response to the conversational input 405. In some instances, a
similar
process to that described in FIGS. 2A-B can be used to identify particular
tokens
associated with the response, where the syntax and semantic generation module
430
generates those tokens based on the substance of the response identified by
the response =
analysis engine 425.
[0091] Once the initial response is determined, the lexical personality filter
module 435 can be used to determine at least one persona-specific
personalization to be
made to the initial response tokens based on a user profile-specific persona
to be applied.
As noted, the social network data 450 acquired by the user profile analysis
engine 427
can be used and compared to the persona library to identify one or more
persona entities
that correspond to the social network activity of the user profile. In some
instances, a
persona response type as identified in the persona matching table 440 may be
associated
with two or more personas. For example, a Motherly persona response type
illustrated in
table 440 corresponds to "Oprah Winfrey." Other implementations may include
additional options for the Motherly persona response type, such as "Julia
Roberts" and
CA 2981261 2017-10-04
"Emma Thompson". The user profile's associated social network data 450 can be
used to
identify a best match of those personas. That best match can be determined by
a direct
like or following of one of the personas, a like or following of a business or
project
associated with one of the personas, and/or a positive or negative reactions
or interaction
associated with one of the personas. By using the social network data, the
lexical
personality filter module 435 can identify a persona recognized and,
hopefully, liked or
trusted by the user profile.
[0092] Each particular persona, or personality library actor as listed in
table 440,
can be associated with at least one persona-related content. In some
instances, that
content can include a persona-specific phrasing library identifying a
particular word or
phrase used by the persona. In some instances, recorded audio of the persona
can be
available such that the voice of the persona can be used to enhance the
response. In some
instances, images and/or video associated with the persona could also be used
to enhance
a presentation of the conversational response 445. For example, an image of or
associated with the persona could be applied to a textual response. If the
persona is
associated with a particular positive or negative recommendation phrase, that
phrase can
be used to supplement or modify a particular response. In some instances, one
or more of
the tokens or portion of the response generated by the syntax and semantic
generation
module 430 can be replaced by the persona-specific content, while in others,
the response
is supplemented with the persona-specific content.
[0093] Once the appropriate persona-specific modification or supplement has
been added to the response, the conversational response 445 can be returned to
the user
401 (or the client device associated with the user 401) by the NLG engine 428.
[0094] FIG. 5 is a flow chart of an example method 500 performed at a system
associated with a conversational interface to transform a responsive output
into a
persona-specific responsive output based on the corresponding user profile. It
will be
understood that method 500 and related methods may be performed, for example,
by any
suitable system, environment, software, and hardware, or a combination of
systems,
environments, software, and hardware, as appropriate. For example, a system
comprising
a communications module, at least one memory storing instructions and other
required
36
CA 2981261 2017-10-04
data, and at least one hardware processor interoperably coupled to the at
least one
memory and the communications module can be used to execute method 500. In
some
implementations, the method 500 and related methods are executed by one or
more
components of the system 100 described above with respect to FIG. 1, or the
components
described in FIG. 4.
[0095] At 505, a first signal including a conversational input associated with
interactions with a conversational interface are received via a communications
module.
The conversational input can include or be associated with an identification
of a
particular user profile associated with the interaction. In some instances,
the identifier
may be included in the conversational input or associated through metadata or
another
suitable manner. In some instances, the conversational input may be received
at a
specific endpoint associated with a particular conversational interface or
related
application, such that the analysis of the conversational input is based on
the use of the
particular conversational interface or related application. Examples may
include a digital
assistant or a particular website, where the conversational input is to be
responded to
based on the information obtainable by or related to that particular interface
or
application.
[0096] As previously noted, the conversational input may be or may include
text-
based input, auditory input (e.g., an audio file), and/or video or image
input. The
conversational input can be submitted by a user or agent associated with a
particular user
profile, although for purposes of some implementations of method 500,
identification of
the particular user profile is not needed to perform the personalization
process. In some
instances, the received conversational input can represent a particular query,
question, or
interactive communication, where the conversational output represents a query
response,
an answer to the question, or a response to the interactive communication.
[0097] At 510, the received conversational input is analyzed via a natural
language processing engine or service to determine an intent associated with
the input
and a personality input type based on the input. In some instances,
determining the
personality input type may include determining a set of lexical personality
scores based
on the content of the received input. The intent associated with the input can
be
37
CA 2981261 2017-10-04
determined based on any suitable factor, and can be used to determine a
meaning of a
particular request associated with the conversational input. The intent may be
used to
determine a particular response to be generated. The lexical personality
scores can be
used to identify various measures of the received conversational input, and
can be
compared to a table or existing set of measures that, when combined, are used
to identify
a particular personality input type. In some instances, two or more
personality input
types may be associated with a single conversational input, such as "angry"
and "scared."
[0098] At 515, a persona response type can be identified, where the persona
response type is associated with or determined based on the determined
personality input
type. In some instances, particular personality input types can be mapped to
one or more
different persona response types, such that once the personality input type is
identified
the mapping can be considered to determine the corresponding persona response
type. In
some instances, two or more persona types may be associated with the
particular
personality input type. In some instances, the persona type may be changed
based on a
response to be returned. For instance, if the conversational interface is
associated with a
financial analysis system, and the conversational input's intent is to ask
whether a
particular transaction should be completed, the response may be a
recommendation
against the transaction. In such an instance, the persona response type may be
associated
with a relatively stricter or more serious persona capable of providing
additional
emphasis to the user that the transaction should not be completed. In the
previous
example, for instance, the Motherly persona response type can be modified to
Strong
based on the response content.
[0099] At 520, a set of response content responsive to the determined intent
of the
received conversational input can be determined. As noted, the operations of
520 may be
performed before those of 515 in some instances. Alternatively, after the
operations of
520, a determination may be made to determine whether to adjust the identified
persona
response type.
[0100] At 525, a particular persona associated with the particular user
profile can
be identified from a plurality of potential personas based, at least in part,
on the social
network activity information associated with the particular user profile. The
particular
38
CA 2981261 2017-10-04
persona identified can correspond to, match, or be associated with the
identified persona
response type (of 515). In some instances, that particular persona may be an
exact match
to the identified persona response type, while in others, the particular
persona may be
similar to, but not identical to, the identified persona response type. As
noted, the
particular persona may be identified based on one or more social network
preferences,
likes, or other indications as they are related to the entity or person
associated with the
particular persona. Each of the personas from the plurality of potential
personas can be
associated with a set of persona-related content. The content may include
persona-
specific words or phrases, such as a catch phrase or other words commonly
associated
with the persona. In some instances, those words or phrases may be synonyms of
particular words, phrases, or tokens included in the response content, and can
replace one
or more tokens from the response content. In other instances, the persona-
specific words
or phrases can be used to supplement and/or otherwise add to the response
content, such
as by using those words as an emphasis to the underlying response. In some
instances,
the persona-related content may include one or more embellishments or non-
response
content, such as an image or video of the person associated with the person.
In some
instances, the persona-related content may include a set of voice prompts or
content that
allows at least a portion of the conversational response to be presented in
the voice of the
particular persona.
[0101] At 530, the set of response content is modified, or enhanced, using at
least
a portion of the persona-related content associated with the particular
persona. The result
of the modification or enhancement is a persona-associated or persona-specific
response.
As noted, the modification may be a replacement of at least a portion of the
response
content with persona-related content, or the modification may be a supplement
to the
response content with additional persona-related content without removing the
initially
determined response content.
[0102] At 535, a second signal including the persona-associated version of the
set
of response content is transmitted, via the communications module, to the
conversational
interface associated with the received first signal. In some instances, that
interface is
associated with a device associated with the particular user profile.
39
CA 2981261 2017-10-04
[0103] The preceding figures and accompanying description illustrate example
processes and computer-implementable techniques. But system 100 (or its
software or
other components) contemplates using, implementing, or executing any suitable
technique for performing these and other tasks. It will be understood that
these processes
are for illustration purposes only and that the described or similar
techniques may be
performed at any appropriate time, including concurrently, individually, or in
combination. In addition, many of the operations in these processes may take
place
simultaneously, concurrently, and/or in different orders than as shown.
Moreover, the
described systems and flows may use processes and/or components with or
performing
additional operations, fewer operations, and/or different operations, so long
as the
methods and systems remain appropriate.
[0104] In other words, although this disclosure has been described in terms of
certain embodiments and generally associated methods, alterations and
permutations of
these embodiments and methods will be apparent to those skilled in the art.
Accordingly,
the above description of example embodiments does not define or constrain this
disclosure. Other changes, substitutions, and alterations are also possible
without
departing from the spirit and scope of this disclosure.
CA 2981261 2017-10-04
