Note: Descriptions are shown in the official language in which they were submitted.
INTELLIGENTLY QUESTIONING AND ANSWERING METHOD, DEVICE,
COMPUTER EQUIPMENT AND STORAGE MEDIUM
BACKGROUND OF THE INVENTION
Technical Field
[0001] The present invention relates to the field of data processing
technology, and more
particularly to an intelligently questioning and answering method, and
corresponding
device, computer equipment and storage medium.
Description of Related Art
[0002] With the rapid development of financial businesses, customer service
departments put
more and more demand on personnel. Problems processed everyday by customer
service
personnel are repetitive and mechanical works to these service personnel as
both the
problems for which advices are sought by users and utterances responded by the
service
personnel are essentially fixed or similar, and huge manpower expenditure has
to be
additionally provided therefor.
[0003] Over the recent years, with the advent of artificial intelligence, many
manpower
consuming works can be realized by computers, the development of artificial
intelligence
has not only become a hotspot of the scientific circle, but has also been the
pursuit of
various intemet companies. From the perspective of company development,
utilization of
artificial intelligence technology to aid or even to replace human works not
only
economizes on the cost but is also a progress of informatization and
intellectualization.
For example, the robot Q&A system can superbly help customer service personnel
in their
work, and consummate Q&A robots can give quick and precise answers to
questions put
forth by users. However, currently available commercial customer service
robots are
1
Date Recue/Date Received 2022-03-22
mostly based on the knowledge base and single-round dialogues lacking
contextual
association, are rather low in working efficiency, and not so high in the
precision in
answering questions, whereby user experience is rendered mediocre.
[0004] Accordingly, there is an urgent need to propose a novel intelligently
questioning and
answering method, so as to address the above problems.
SUMMARY OF THE INVENTION
[0005] In order to deal with problems pending in the state of the art,
embodiments of the present
invention provide an intelligently questioning and answering method, and
corresponding
device, computer equipment and storage medium, so as to overcome problems
prevailing
in prior-art intelligent Q&A technology in which contextual association is
lacking and
precision in answering questions is relatively low.
[0006] To solve one or more of the aforementioned technical problem(s), the
present invention
employs the following technical solutions.
[0007] According to the first aspect, there is provided an intelligently
questioning and answering
method that comprises the following steps:
[0008] performing word-segmentation processing on a received questioning
statement sent from
a current user, and obtaining a word-segmentation result of the questioning
statement;
[0009] determining an operation scenario of the questioning statement
according to a preset
decision model and a preset rule;
[0010] employing, when the operation scenario is a Q&A scenario, a preset
classification model
and the word-segmentation result to recognize a current user intention and a
current
business scenario of the questioning statement, storing a recognition result
in association
with the current user, and determining a target calculating rule of the
questioning
statement according to the recognition result;
2
Date Recue/Date Received 2022-03-22
[0011] performing corresponding calculation on the word-segmentation result of
the questioning
statement according to the target calculating rule, and obtaining a
calculation result; and
[0012] generating result data of a preset format according to the calculation
result, so as to
facilitate check by the current user.
[0013] Further, the step of performing corresponding calculation on the word-
segmentation
result of the questioning statement according to the target calculating rule,
and obtaining
a calculation result includes:
[0014] calculating a similarity between the questioning statement and
candidate sentences in a
preset Q&A library according to the word-segmentation result.
[0015] Further, when the recognition result does not contain any current
business scenario but
contains the current user intention, the method further comprises:
[0016] enquiring whether the current user has any historical business
scenario, if yes, judging
whether the historical business scenario is related to the current user
intention.
[0017] Further, when the recognition result does not contain any current
business scenario but
contains the current user intention, and the current user does not have any
historical
business scenario or the historical business scenario is not related to the
current user
intention, the step of determining a target calculating rule of the
questioning statement
according to the recognition result includes:
[0018] employing a preset map matching model and the word-segmentation result
to retrieve out
a plurality of candidate business scenarios related to the current user
intention and to feed
back the same to the current user for selection, and determining the target
calculating rule
of the questioning statement according to the candidate business scenario
selected by the
current user.
[0019] Further, when the recognition result contains the current business
scenario and contains
the current user intention, or when the recognition result does not contain
any current
3
Date Recue/Date Received 2022-03-22
business scenario but contains the current user intention, and the historical
business
scenario is related to the current user intention, the step of determining a
target calculating
rule of the questioning statement according to the recognition result
includes:
[0020] determining the target calculating rule as calculating a similarity
between the word-
segmentation result of the questioning statement and candidate sentences in
the preset
Q&A library.
[0021] Further, when the recognition result contains the current business
scenario but does not
contain any current user intention, the method further comprises:
[0022] employing a preset map matching model and the word-segmentation result
to retrieve out
a plurality of candidate user intentions related to the current business
scenario and to feed
back the same to the current user for selection, and determining the target
calculating rule
of the questioning statement according to the candidate user intention
selected by the
current user.
[0023] Further, when the recognition result does not contain any current
business scenario and
does not contain any current user intention, the method further comprises:
[0024] pushing preset hotspot questions to the current user for selection.
[0025] According to the second aspect, there is provided an intelligently
questioning and
answering device that comprises:
[0026] a data processing module, for performing word-segmentation processing
on a received
questioning statement sent from a current user, and obtaining a word-
segmentation result
of the questioning statement;
[0027] a first recognizing module, for determining an operation scenario of
the questioning
statement according to a preset decision model and a preset rule;
[0028] a second recognizing module, for employing, when the operation scenario
is a Q&A
scenario, a preset classification model and the word-segmentation result to
recognize a
current user intention and a current business scenario of the questioning
statement, and
4
Date Recue/Date Received 2022-03-22
determining a target calculating rule of the questioning statement according
to the
recognition result;
[0029] a data calculating module, for performing corresponding calculation on
the word-
segmentation result of the questioning statement according to the target
calculating rule,
and obtaining a calculation result; and
[0030] a result outputting module, for generating result data of a preset
format according to the
calculation result, so as to facilitate check by the current user.
[0031] According to the third aspect, there is provided a computer equipment
that comprises a
memory, a processor and a computer program stored on the memory and operable
on the
processor, and the following steps are realized when the processor executes
the computer
program:
[0032] performing word-segmentation processing on a received questioning
statement sent from
a current user, and obtaining a word-segmentation result of the questioning
statement;
[0033] determining an operation scenario of the questioning statement
according to a preset
decision model and a preset rule;
[0034] employing, when the operation scenario is a Q&A scenario, a preset
classification model
and the word-segmentation result to recognize a current user intention and a
current
business scenario of the questioning statement, storing a recognition result
in association
with the current user, and determining a target calculating rule of the
questioning
statement according to the recognition result;
[0035] performing corresponding calculation on the word-segmentation result of
the questioning
statement according to the target calculating rule, and obtaining a
calculation result; and
[0036] generating result data of a preset format according to the calculation
result, so as to
facilitate check by the current user.
[0037] According to the fourth aspect, there is provided a computer-readable
storage medium
storing a computer program thereon, and the following steps are realized when
the
computer program is executed by a processor:
Date Recue/Date Received 2022-03-22
[0038] performing word-segmentation processing on a received questioning
statement sent from
a current user, and obtaining a word-segmentation result of the questioning
statement;
[0039] determining an operation scenario of the questioning statement
according to a preset
decision model and a preset rule;
[0040] employing, when the operation scenario is a Q&A scenario, a preset
classification model
and the word-segmentation result to recognize a current user intention and a
current
business scenario of the questioning statement, storing a recognition result
in association
with the current user, and determining a target calculating rule of the
questioning
statement according to the recognition result;
[0041] performing corresponding calculation on the word-segmentation result of
the questioning
statement according to the target calculating rule, and obtaining a
calculation result; and
[0042] generating result data of a preset format according to the calculation
result, so as to
facilitate check by the current user.
[0043] The technical solutions provided by the embodiments of the present
invention bring about
the following advantageous effects.
[0044] In the intelligently questioning and answering method, and
corresponding device,
computer equipment and storage medium provided by the embodiments of the
present
invention, by performing word-segmentation processing on a received
questioning
statement sent from a current user, obtaining a word-segmentation result of
the
questioning statement, determining an operation scenario of the questioning
statement
according to a preset decision model and a preset rule, employing, when the
operation
scenario is a Q&A scenario, a preset classification model and the word-
segmentation
result to recognize a current user intention and a current business scenario
of the
questioning statement, storing a recognition result in association with the
current user,
determining a target calculating rule of the questioning statement according
to the
recognition result, performing corresponding calculation on the word-
segmentation result
of the questioning statement according to the target calculating rule,
obtaining a
6
Date Recue/Date Received 2022-03-22
calculation result, and generating result data of a preset format according to
the
calculation result, so as to facilitate check by the current user, and by
recognizing the
business scenario and the user intention of the questioning statement
respectively,
precision in answering the question is enhanced, and user experience hence is
enhanced.
[0045] In the intelligently questioning and answering method, and
corresponding device,
computer equipment and storage medium provided by the embodiments of the
present
invention, after the current user intention and the current business scenario
of the
questioning statement have been recognized by employing the preset
classification model
and the word-segmentation result, the recognition result is stored in
association with the
current user, so that, when a subsequent questioning statement sent from the
user for
seeking advice for a question lacks business scenario, the stored recognition
result is
invoked to serve as reference, so as to further enhance precision in answering
the question.
BRIEF DESCRIPTION OF THE DRAWINGS
[0046] To more clearly describe the technical solutions in the embodiments of
the present
invention, drawings required to be used in the description of the embodiments
will be
briefly introduced below. Apparently, the drawings introduced below are merely
directed
to some embodiments of the present invention, while it is possible for persons
ordinarily
skilled in the art to acquire other drawings based on these drawings without
spending
creative effort in the process.
[0047] Fig. 1 is a flowchart illustrating the intelligently questioning and
answering method
according to an exemplary embodiment;
[0048] Fig. 2 is a view schematically illustrating the structure of the
intelligently questioning and
answering device according to an exemplary embodiment; and
7
Date Recue/Date Received 2022-03-22
[0049] Fig. 3 is a view schematically illustrating the internal structure of
the computer equipment
according to an exemplary embodiment.
DETAILED DESCRIPTION OF THE INVENTION
[0050] In order to make more lucid and clear the objectives, technical
solutions and advantages
of the present invention, the technical solutions in the embodiments of the
present
invention will be more clearly and comprehensively described below with
reference to
the accompanying drawings in the embodiments of the present invention.
Apparently, the
embodiments as described are merely partial, rather than the entire,
embodiments of the
present invention. All other embodiments obtainable by persons ordinarily
skilled in the
art on the basis of the embodiments in the present invention without spending
creative
effort shall all fall within the protection scope of the present invention.
[0051] As noted in the Description of Related Art, in order to catch up with
the development of
financial businesses, to reduce manpower cost, and to enhance customer service
efficiency and user satisfaction, it is usually required for intelligent
customer service
robots to cooperatively work with the human customer service. However,
currently
available commercial customer service robots are mostly based on the knowledge
base
and single-round dialogues lacking contextual association, are relatively low
in the
precision in answering questions, whereby working efficiency and customer
service
experience are rendered mediocre.
[0052] To address the aforementioned problems, in the embodiments of the
present invention is
created a complete financial business intelligently questioning and answering
robot
system having such functions as intention recognition, contextual association,
map
retrieval and matching of similar questions by previously combining such
technologies
as the natural language processing technology, machine learning technology,
deep
learning technology, and knowledge map technology, etc. During specific
8
Date Recue/Date Received 2022-03-22
implementation, the intelligently questioning and answering robot system can
be created
on the basis of a Q&A knowledge base and a knowledge map, as a preferred
examples,
the knowledge base is embodied as an ES knowledge base, and construction of
the
knowledge base includes a financially professional Q&A library, a chitchat Q&A
library,
a hotspot questions library, a user log library, and a scenario-intention
relations library.
The knowledge map can be constructed as a financial Q&A map according to data
in the
financially professional Q&A library.
[0053] Based on such a system can be realized the intelligently questioning
and answering
method proposed by an embodiment of the present invention, with respect to the
operation scenario of a questioning statement sent from the user being a Q&A
scenario,
this method further recognizes the current user intention and the current
business scenario
of the questioning statement, on the one hand, a target calculating rule is
determined in
conjunction with the recognized current user intention and current business
scenario to
perform corresponding calculation on the questioning statement, result data of
a preset
format is thereafter generated according to the calculation result and fed
back to the
current user, so that precision in answering the question is enhanced, on the
other hand,
the recognition result is stored in association with the current user, so
that, while the user
is carrying out a round-to-round dialogue, when a subsequently sent
questioning
statement lacks business scenario, the stored recognition result is invoked to
serve as
reference, so as to further enhance precision in answering the question.
[0054] Fig. 1 is a flowchart illustrating the intelligently questioning and
answering method
according to an exemplary embodiment, with reference to Fig. 1, the method
comprises
the following steps.
[0055] 51 - performing word-segmentation processing on a received questioning
statement sent
from a current user, and obtaining a word-segmentation result of the
questioning
statement.
9
Date Recue/Date Received 2022-03-22
[0056] Specifically, it is usually required to somehow preprocess in advance
the data that enters
the intelligently questioning and answering robot system, so as to enhance the
precision
of subsequent calculation. In the embodiments of the present invention,
besides including
performing word-segmentation processing on the received questioning statement
to
obtain the word-segmentation result in the questioning statement, the
preprocessing
operation can further include such operations as character purification, word
segmentation, and correction, etc., to which no restriction is made in this
context, and it
is possible for the user to set as practically required.
[0057] Taking for example an intelligently questioning and answering robot
system of the
financial business, a self-defined dictionary pertaining to the field of
financial business
can be constructed in advance to perform word-segmentation processing on the
questioning statement sent from the current user, a self-defined correction
dictionary
pertaining to the field of financial business can also be constructed in
advance to perform
correction processing on the questioning statement sent from the current user.
[0058] S2 - determining an operation scenario of the questioning statement
according to a preset
decision model and a preset rule.
[0059] Specifically, usually speaking, questions for which advices are sought
by users are
classified into plural operation scenarios. In the embodiments of the present
invention,
after the word-segmentation result of the questioning statement has been
extracted, a
preset decision model and a preset rule will be employed to recognize the
keyword to
determine the specific operation scenario to which the questioning statement
corresponds,
namely to classify the user's question to the specific operation scenario for
directed
solution. The preset rule includes, but is not limited to, a keyword rule, for
example,
different keywords are set in advance for different operation scenarios, the
word-
segmentation result is matched with preset keywords, once the matching result
satisfies
Date Recue/Date Received 2022-03-22
the preset requirement, the questioning statement is classified to the
operation scenario to
which the matched keyword corresponds. During specific implementation, the
preset
decision model can be trained and obtained in advance by the use of the
keyword rule
and a machine learning model. As should be noted here, in the embodiments of
the present
invention, the business scenario includes, but is not limited to, Capricious
Loan,
Capricious Payment, and Change Treasure, etc., and the user intention
includes, but is not
limited to, repaying, cancelling, and changing password, etc.
[0060] Likewise taking the financial business for example, in the embodiments
of the present
invention, the operation scenario can be classified in advance mainly into the
following
types: a manual scenario, a chitchat scenario, an order scenario, a self-help
operation, a
Q&A scenario, and a similar question clicking scenario, etc. The manual
scenario is used
to solve problems proposed on the user's own initiative to be transferred for
manual
solution or recognized by the system according to a preset rule to be
necessarily manually
solved, and the keywords as well as the intentions can be configured by the
foreground
service personnel as keywords and intentions involved and processed in the
transfer to
manual solution. The chitchat scenario makes use of a colossal chitchat
knowledge base
to solve such chitchat problems as concerning greeting, praising, and
expressing involved
in the process of seeking advice by the user. The order scenario solves
problems possibly
encountered by the user during the process of payment, and subsequently
invokes recent
orders of the user to present to the user. The self-help operation solves some
problems to
be completed by a series of operations by providing operation links. What the
Q&A
system returns after one round of Q&A may not be the answer, but similar
questions for
the user to select, these similar questions are already existent in the
knowledge base, and
their answers can be returned by direct retrieval after clicking of the user,
what the similar
question clicking scenario solves are precisely these types of questions.
[0061] When the preset decision model recognizes that the operation scenario
of the questioning
statement is such a scenario as manual, order, or self-help operation, the
outputting
11
Date Recue/Date Received 2022-03-22
module is directly entered, and respective identifications are output; when
the preset
decision model recognizes that the operation scenario of the questioning
statement is a
chitchat scenario, the knowledge base is entered to match out the answer; when
the preset
decision model recognizes that the operation scenario of the questioning
statement is a
Q&A scenario, the financial robot Q&A scenario is entered, and the subsequent
flow is
continued.
[0062] S3 - employing, when the operation scenario is a Q&A scenario, a preset
classification
model and the word-segmentation result to recognize a current user intention
and a
current business scenario of the questioning statement, storing a recognition
result in
association with the current user, and determining a target calculating rule
of the
questioning statement according to the recognition result.
[0063] Specifically, the method provided by the embodiment of the present
invention is proposed
to mainly solve questions sent from the user under the Q&A scenario. Since a
question
usually basically contains two key elements, namely business scenario and user
intention,
when it is recognized that the operation scenario of the questioning statement
is a Q&A
scenario, it is required to further recognize the user intention and the
specific business
scenario of the questioning statement, and a target calculating rule of the
questioning
statement is thereafter determined according to the recognition result.
[0064] Specifically, the intention recognizing module is the most core part of
the Q&A robot,
and intention recognition of the user's question directly affects the Q&A
effect, in the
embodiments of the present invention, two independent deep learning
classifiers are used
for intention recognition, of which one is used to recognize the business
scenario of the
question, and the other one is used to recognize the intention involved in the
question.
During specific implementation, a confusion threshold and certain specific
rules can be
used to judge whether the results of the classifiers are believable, once a
preset condition
is met, it is considered that the classification result is confused, that is,
the classifier
12
Date Recue/Date Received 2022-03-22
cannot extract the correct business scenario or user intention, conversely,
the result is not
confused, that is, the classifier can extract the correct business scenario or
user intention.
[0065] Specifically, when the user seeks advice for a question, the user
mostly directly speaks
out the intention, but the scenario is lacking, accordingly, when the
intention recognizing
module is passed during the process of multiple rounds of Q&A of the user, the
scenario
recognized by the current model will be stored in association with the current
user to
serve as reference data for the next Q&A scenario.
[0066] S4 ¨ performing corresponding calculation on the word-segmentation
result of the
questioning statement according to the target calculating rule, and obtaining
a calculation
result.
[0067] Specifically, the recognition result obtained by employing a preset
classification model
and the word-segmentation result to recognize a current user intention and a
current
business scenario of the questioning statement may subsume plural
circumstances, in the
embodiments of the present invention, different follow-up operations are set
with respect
to the different recognition results, so as to enhance the precision in
answering the
question ¨ the specific modes of execution will be enunciated later on a one-
by-one basis.
[0068] S5 ¨ generating result data of a preset format according to the
calculation result, so as to
facilitate check by the current user.
[0069] Specifically, in order to enhance user experience, in the embodiments
of the present
invention, after the calculation result has been obtained, the calculation
result can be
sorted into various forms required by the frontend to be fed back to the
current user. For
instance, with respect to a question with a specific special identification as
returned by
the first recognizing module, the question is sorted into a result with the
special
identification and returned; with respect to a request for hotspot questions
as returned by
13
Date Recue/Date Received 2022-03-22
the second recognizing module, the hotspot questions are retrieved and sorted
into the
form of a link of similar questions to be returned; with respect to plural
options as returned
by the map matching module, several top-ranking options in terms of
probabilities are
selected and sorted into slot candidate options to be returned; with respect
to a calculation
result as returned by the data calculating module, if the result is an answer,
the answer is
sorted and returned, if the result is a plurality of similar questions, these
are sorted into
the form of a link of similar questions and returned; with respect to an
answer to a chitchat
question as returned by the knowledge base matching module, the answer is
sorted and
returned, and with respect to similar questions as returned by the data
calculating module,
the similar questions are sorted into the form of a link of similar questions
and returned -
-- to which no enumeration is made on a one-by-one basis in this context.
[0070] As a preferred mode of execution in the embodiments of the present
invention, the step
of performing corresponding calculation on the word-segmentation result of the
questioning statement according to the target calculating rule, and obtaining
a calculation
result includes:
[0071] calculating a similarity between the questioning statement and
candidate sentences in a
preset Q&A library according to the word-segmentation result.
[0072] Specifically, in the case the questioning statement sent from the
current user has the two
key elements as the complete business scenario and user intention, the
questioning
statement will enter the data calculating module for calculation. As a
preferred example
in the embodiments of the present invention, the calculation of the word-
segmentation
result includes similarity calculation. During specific implementation, a
similarity
between the questioning statement of the current user and questions (namely
candidate
sentences) in a preset Q&A library (that includes, but is not limited to, a
financially
professional Q&A library) is calculated through the word-segmentation result.
The
similarity model can be modified on the basis of the Alibaba text matching
model (SETM),
to conform to the data in the financial Q&A library. What the similarity model
outputs
14
Date Recue/Date Received 2022-03-22
are candidate sentences and their similarity scores.
[0073] As a preferred mode of execution in the embodiments of the present
invention, two
screening thresholds can be set in advance, of which one is a standard
question threshold,
the question (namely a candidate sentence) satisfying this threshold is
determined as the
questioning statement sent from the current user, and the answer to this
question is taken
to serve as the calculation result; the other one is a similar question
threshold, the question
(namely a candidate sentence) satisfying this threshold can serve as a
candidate similar
question of the questioning statement sent from the current user, and the
candidate similar
question is taken to serve as the calculation result. Put in other words, the
calculation
result returned by the data calculating module is classified into two
circumstances, one is
the answer to the question, and the other one is a plurality of similar
questions whose
similarities have been sequenced.
[0074] As a preferred mode of execution in the embodiments of the present
invention, when the
recognition result does not contain any current business scenario but contains
the current
user intention, the method further comprises:
[0075] enquiring whether the current user has any historical business
scenario, if yes, judging
whether the historical business scenario is related to the current user
intention.
[0076] Specifically, in order to enhance the Q&A precision, in the embodiments
of the present
invention, when it is recognized that the questioning statement sent from the
current user
lacks business scenario, it will be enquired whether the current user has any
historical
business scenario, the historical business scenario here includes, but is not
limited to,
business scenario data contained in the recognition result obtained as the
second
recognizing module recognizes the current user intention and the current
business
scenario of the questioning statement in the process of plural rounds of Q&As
by the
current user. If it is enquired that the current user has an associated
historical business
scenario, it is further judged whether the historical business scenario is
related to the
Date Recue/Date Received 2022-03-22
current user intention.
[0077] As a preferred mode of execution in the embodiments of the present
invention, when the
recognition result does not contain any current business scenario but contains
the current
user intention, and the current user does not have any historical business
scenario or the
historical business scenario is not related to the current user intention, the
step of
determining a target calculating rule of the questioning statement according
to the
recognition result includes:
[0078] employing a preset map matching model and the word-segmentation result
to retrieve out
a plurality of candidate business scenarios related to the current user
intention and to feed
back the same to the current user for selection, and determining the target
calculating rule
of the questioning statement according to the candidate business scenario
selected by the
current user.
[0079] Specifically, when the current user intention is recognized, but no
current business
scenario is recognized or the current user does not have any historical
business scenario,
or when the current user has a historical business scenario but the historical
business
scenario is not related to the current user intension, a preset map matching
model and the
word-segmentation result are employed to retrieve out a plurality of candidate
business
scenarios related to the current user intention, the plural candidate business
scenarios are
subsequently fed back to the current user for selection, and the target
calculating rule of
the keyword is determined according to the candidate business scenario clicked
and
selected by the current user in conjunction with the current user intention
obtained in the
foregoing step. The target calculating rule here includes, but is not limited
to, calculating
the similarity between the keyword and candidate sentences in the preset Q&A
library.
[0080] During specific implementation, taking the financial business for
example, a financial
Q&A knowledge map can be constructed by means of the questions in the
financially
professional Q&A library, with scenarios and intentions of financial questions
serving as
16
Date Recue/Date Received 2022-03-22
nodes of the map. The knowledge map is mainly used to retrieve according to
recognized
business scenarios or user intentions when it is recognized that the
questioning statement
sent from the current user is incomplete in terms of the business scenario or
the user
intention, and to provide plural candidate business scenarios or candidate
user intentions
with higher probabilities and to return the same to the current user for
selection.
[0081] As a preferred example in the embodiments of the present invention, it
is further possible
to take the candidate business scenarios or user intentions as the calculation
result and to
input the same to the outputting module, and to subsequently sort the result
into result
data of a preset format to be fed back to the current user for selection.
[0082] As a preferred mode of execution in the embodiments of the present
invention, when the
recognition result contains the current business scenario and contains the
current user
intention, or when the recognition result does not contain any current
business scenario
but contains the current user intention, and the historical business scenario
is related to
the current user intention, the step of determining a target calculating rule
of the
questioning statement according to the recognition result includes:
[0083] determining the target calculating rule as calculating a similarity
between the word-
segmentation result of the questioning statement and candidate sentences in
the preset
Q&A library.
[0084] Specifically, when the current user intention is recognized and the
current business
scenario is recognized, or when the current user intention is recognized, but
no current
business scenario is recognized but the current user has a historical business
scenario and
the historical business scenario is related to the current user intention, it
is possible to
directly determine the target calculating rule as calculating a similarity
between the word-
segmentation result of the questioning statement and candidate sentences in
the preset
Q&A library.
17
Date Recue/Date Received 2022-03-22
[0085] As a preferred mode of execution in the embodiments of the present
invention, when the
recognition result contains the current business scenario but does not contain
any current
user intention, the method further comprises:
[0086] employing a preset map matching model and the word-segmentation result
to retrieve out
a plurality of candidate user intentions related to the current business
scenario and to feed
back the same to the current user for selection, and determining the target
calculating rule
of the questioning statement according to the candidate user intention
selected by the
current user.
[0087] Specifically, when the current business scenario is recognized but no
current user
intention is recognized, a preset map matching model and the word-segmentation
result
are employed to retrieve out a plurality of candidate user intentions involved
in the current
business scenario, the plural candidate user intentions are subsequently fed
back to the
current user for selection, and the target calculating rule of the keyword is
determined
according to the candidate user intention clicked and selected by the current
user in
conjunction with the aforementioned current business scenario. The target
calculating
rule here includes, but is not limited to, calculating a similarity between
the word-
segmentation result and candidate sentences in the preset Q&A library, namely
calculating the similarity between the questioning statement and candidate
sentences in
the preset Q&A library.
[0088] As a preferred mode of execution in the embodiments of the present
invention, when the
recognition result does not contain any current business scenario and does not
contain
any current user intention, the method further comprises:
[0089] pushing preset hotspot questions to the current user for selection.
[0090] Specifically, when none of the current business scenario and current
user intention
involved in the questioning statement is recognized, it is now impossible to
judge the
question specifically asked by the current user, at this time it is possible
to push preset
18
Date Recue/Date Received 2022-03-22
hotspot questions to the current user for selection. As should be noted here,
the preset
hotspot questions in the embodiments of the present invention are marked in
advance
with their business scenarios and user intentions, once the current user
selects a certain
hotspot question, the answer to this hotspot question is directly fed back to
the current
user.
[0091] As a preferred mode of execution in the embodiments of the present
invention, a
knowledge base is further constructed in advance. When it is recognized that
the operation
scenario of the questioning statement is a chitchat scenario, the answer is
directly
retrieved out through matching of the knowledge base, and the answer is input
to the
result outputting module. Alternatively, after a questioning statement having
the complete
business scenario and user intention has been calculated through the data
calculating
module, there might be the circumstance in which the lowest similar question
threshold
cannot be found. Such a circumstance may be due to the fact that the business
personnel
has updated the financial Q&A library, or that the similarity model of this
questioning
statement can indeed not be recognized. In order to solve such problems,
analysis is
further performed in the embodiments of the present invention by employing ES
retrieval
and an algorithm based on word movement distance (WMD) after the data
calculating
module. The sorting retrieval function carried with the ES is firstly utilized
to retrieve 50
pieces of data returned by the financially professional Q&A library,
similarities between
the user's questions and the returned data are calculated and sorted by means
of the WMD
algorithm, and a preset threshold is invoked to the effect that questions
satisfying this
threshold can be taken as candidate similar questions and input to the result
outputting
module.
[0092] Fig. 2 is a view schematically illustrating the structure of the
intelligently questioning and
answering device according to an exemplary embodiment, with reference to Fig.
2, the
device comprises:
[0093] a data processing module, for performing word-segmentation processing
on a received
19
Date Recue/Date Received 2022-03-22
questioning statement sent from a current user, and obtaining a word-
segmentation result
of the questioning statement;
[0094] a first recognizing module, for determining an operation scenario of
the questioning
statement according to a preset decision model and a preset rule;
[0095] a second recognizing module, for employing, when the operation scenario
is a Q&A
scenario, a preset classification model and the word-segmentation result to
recognize a
current user intention and a current business scenario of the questioning
statement, and
determining a target calculating rule of the questioning statement according
to the
recognition result;
[0096] a data calculating module, for performing corresponding calculation on
the word-
segmentation result of the questioning statement according to the target
calculating rule,
and obtaining a calculation result; and
[0097] a result outputting module, for generating result data of a preset
format according to the
calculation result, so as to facilitate check by the current user.
[0098] As a preferred mode of execution in the embodiments of the present
invention, the data
calculating module is specifically employed for:
[0099] calculating a similarity between the questioning statement and
candidate sentences in a
preset Q&A library according to the word-segmentation result.
[0100] As a preferred mode of execution in the embodiments of the present
invention, the second
recognizing module is further employed for:
[0101] enquiring whether the current user has any historical business
scenario, if yes, judging
whether the historical business scenario is related to the current user
intention.
[0102] As a preferred mode of execution in the embodiments of the present
invention, the device
further comprises:
[0103] a map matching module, for employing a preset map matching model and
the word-
segmentation result to retrieve out a plurality of candidate business
scenarios related to
Date Recue/Date Received 2022-03-22
the current user intention and to feed back the same to the current user for
selection;
[0104] the second recognizing module is specifically employed for determining
the target
calculating rule of the questioning statement according to the candidate
business scenario
selected by the current user.
[0105] As a preferred mode of execution in the embodiments of the present
invention, the second
recognizing module is specifically employed for:
[0106] determining the target calculating rule as calculating a similarity
between the word-
segmentation result of the questioning statement and candidate sentences in
the preset
Q&A library.
[0107] As a preferred mode of execution in the embodiments of the present
invention, the map
matching module is further employed for:
[0108] employing a preset map matching model and the word-segmentation result
to retrieve out
a plurality of candidate user intentions related to the current business
scenario and to feed
back the same to the current user for selection; and
[0109] the second recognizing module is specifically employed for determining
the target
calculating rule of the questioning statement according to the candidate user
intention
selected by the current user.
[0110] As a preferred mode of execution in the embodiments of the present
invention, the result
outputting module is further employed for:
[0111] pushing preset hotspot questions to the current user for selection.
[0112] Fig. 3 is a view schematically illustrating the internal structure of
the computer equipment
according to an exemplary embodiment, with reference to Fig. 3, the computer
equipment
comprises a processor, a memory, and a network interface connected to each
other via a
system bus. The processor of the computer equipment is employed to provide
computing
and controlling capabilities. The memory of the computer equipment includes a
21
Date Recue/Date Received 2022-03-22
nonvolatile storage medium and an internal memory. The nonvolatile storage
medium
stores therein an operating system, a computer program and a database. The
internal
memory provides environment for the running of the operating system and the
computer
program in the nonvolatile storage medium. The network interface of the
computer
equipment is employed to connect to an external terminal via network for
communication.
The computer program realizes a method of optimizing an execution plan when it
is
executed by a processor.
[0113] As understandable to persons skilled in the art, the structure
illustrated in Fig. 3 is merely
a block diagram of partial structure relevant to the solution of the present
invention, and
does not constitute any restriction to the computer equipment on which the
solution of
the present invention is applied, as the specific computer equipment may
comprise
component parts that are more than or less than those illustrated in Fig. 3,
or may combine
certain component parts, or may have different layout of component parts.
[0114] As a preferred mode of execution in the embodiments of the present
invention, the
computer equipment comprises a memory, a processor and a computer program
stored
on the memory and operable on the processor, and the following steps are
realized when
the processor executes the computer program:
[0115] performing word-segmentation processing on a received questioning
statement sent from
a current user, and obtaining a word-segmentation result of the questioning
statement;
[0116] determining an operation scenario of the questioning statement
according to a preset
decision model and a preset rule;
[0117] employing, when the operation scenario is a Q&A scenario, a preset
classification model
and the word-segmentation result to recognize a current user intention and a
current
business scenario of the questioning statement, storing a recognition result
in association
with the current user, and determining a target calculating rule of the
questioning
statement according to the recognition result;
[0118] performing corresponding calculation on the word-segmentation result of
the questioning
22
Date Recue/Date Received 2022-03-22
statement according to the target calculating rule, and obtaining a
calculation result; and
[0119] generating result data of a preset format according to the calculation
result, so as to
facilitate check by the current user.
[0120] As a preferred mode of execution in the embodiments of the present
invention, when the
processor executes the computer program, the following step is further
realized:
[0121] calculating a similarity between the questioning statement and
candidate sentences in a
preset Q&A library according to the word-segmentation result.
[0122] As a preferred mode of execution in the embodiments of the present
invention, when the
processor executes the computer program, the following step is further
realized:
[0123] enquiring whether the current user has any historical business
scenario, if yes, judging
whether the historical business scenario is related to the current user
intention.
[0124] As a preferred mode of execution in the embodiments of the present
invention, when the
processor executes the computer program, the following steps are further
realized:
[0125] employing a preset map matching model and the word-segmentation result
to retrieve out
a plurality of candidate business scenarios related to the current user
intention and to feed
back the same to the current user for selection, and determining the target
calculating rule
of the questioning statement according to the candidate business scenario
selected by the
current user.
[0126] As a preferred mode of execution in the embodiments of the present
invention, when the
processor executes the computer program, the following step is further
realized:
[0127] determining the target calculating rule as calculating a similarity
between the word-
segmentation result of the questioning statement and candidate sentences in
the preset
Q&A library.
[0128] As a preferred mode of execution in the embodiments of the present
invention, when the
23
Date Recue/Date Received 2022-03-22
processor executes the computer program, the following steps are further
realized:
[0129] employing a preset map matching model and the word-segmentation result
to retrieve out
a plurality of candidate user intentions related to the current business
scenario and to feed
back the same to the current user for selection, and determining the target
calculating rule
of the questioning statement according to the candidate user intention
selected by the
current user.
[0130] As a preferred mode of execution in the embodiments of the present
invention, when the
processor executes the computer program, the following step is further
realized:
[0131] pushing preset hotspot questions to the current user for selection.
[0132] In the embodiments of the present invention, there is further provided
a computer-
readable storage medium storing thereon a computer program, and the following
steps
are realized when the computer program is executed by a processor:
[0133] performing word-segmentation processing on a received questioning
statement sent from
a current user, and obtaining a word-segmentation result of the questioning
statement;
[0134] determining an operation scenario of the questioning statement
according to a preset
decision model and a preset rule;
[0135] employing, when the operation scenario is a Q&A scenario, a preset
classification model
and the word-segmentation result to recognize a current user intention and a
current
business scenario of the questioning statement, storing a recognition result
in association
with the current user, and determining a target calculating rule of the
questioning
statement according to the recognition result;
[0136] performing corresponding calculation on the word-segmentation result of
the questioning
statement according to the target calculating rule, and obtaining a
calculation result; and
[0137] generating result data of a preset format according to the calculation
result, so as to
facilitate check by the current user.
[0138] As a preferred mode of execution in the embodiments of the present
invention, when the
24
Date Recue/Date Received 2022-03-22
computer program is executed by a processor, the following step is further
realized:
[0139] calculating a similarity between the questioning statement and
candidate sentences in a
preset Q&A library according to the word-segmentation result.
[0140] As a preferred mode of execution in the embodiments of the present
invention, when the
computer program is executed by a processor, the following step is further
realized:
[0141] enquiring whether the current user has any historical business
scenario, if yes, judging
whether the historical business scenario is related to the current user
intention.
[0142] As a preferred mode of execution in the embodiments of the present
invention, when the
computer program is executed by a processor, the following steps are further
realized:
[0143] employing a preset map matching model and the word-segmentation result
to retrieve out
a plurality of candidate business scenarios related to the current user
intention and to feed
back the same to the current user for selection, and determining the target
calculating rule
of the questioning statement according to the candidate business scenario
selected by the
current user.
[0144] As a preferred mode of execution in the embodiments of the present
invention, when the
computer program is executed by a processor, the following step is further
realized:
[0145] determining the target calculating rule as calculating a similarity
between the word-
segmentation result of the questioning statement and candidate sentences in
the preset
Q&A library.
[0146] As a preferred mode of execution in the embodiments of the present
invention, when the
computer program is executed by a processor, the following steps are further
realized:
[0147] employing a preset map matching model and the word-segmentation result
to retrieve out
a plurality of candidate user intentions related to the current business
scenario and to feed
back the same to the current user for selection, and determining the target
calculating rule
of the questioning statement according to the candidate user intention
selected by the
Date Recue/Date Received 2022-03-22
current user.
[0148] As a preferred mode of execution in the embodiments of the present
invention, when the
computer program is executed by a processor, the following step is further
realized:
[0149] pushing preset hotspot questions to the current user for selection.
[0150] In summary, the technical solutions provided by the embodiments of the
present invention
bring about the following advantageous effects.
[0151] In the intelligently questioning and answering method, and
corresponding device,
computer equipment and storage medium provided by the embodiments of the
present
invention, by performing word-segmentation processing on a received
questioning
statement sent from a current user, obtaining a word-segmentation result of
the
questioning statement, determining an operation scenario of the questioning
statement
according to a preset decision model and a preset rule, employing, when the
operation
scenario is a Q&A scenario, a preset classification model and the word-
segmentation
result to recognize a current user intention and a current business scenario
of the
questioning statement, storing a recognition result in association with the
current user,
determining a target calculating rule of the questioning statement according
to the
recognition result, performing corresponding calculation on the word-
segmentation result
of the questioning statement according to the target calculating rule,
obtaining a
calculation result, and generating result data of a preset format according to
the
calculation result, so as to facilitate check by the current user, and by
taking the business
scenario and the user intention of the questioning statement into overall
consideration,
precision in answering the question is enhanced, and user experience hence is
enhanced.
[0152] In the intelligently questioning and answering method, and
corresponding device,
computer equipment and storage medium provided by the embodiments of the
present
invention, after the current user intention and the current business scenario
of the
26
Date Recue/Date Received 2022-03-22
questioning statement have been recognized by employing the preset
classification model
and the word-segmentation result, the recognition result is stored in
association with the
current user, so that, when a subsequent questioning statement sent from the
user for
seeking advice for a question lacks business scenario, the stored recognition
result is
invoked to serve as reference, so as to further enhance precision in answering
the question.
[0153] As should be noted, when the intelligently questioning and answering
device provided by
the aforementioned embodiment triggers a questioning and answering business,
the
division into the aforementioned various functional modules is merely by way
of example,
while it is possible, in actual application, to base on requirements to assign
the functions
to different functional modules for completion, that is to say, to divide the
internal
structure of the device into different functional modules to complete the
entire or partial
functions described above. In addition, the intelligently questioning and
answering device
provided by the aforementioned embodiment pertains to the same conception as
the
intelligently questioning and answering method provided by the method
embodiment,
that is to say, the device is based on the intelligently questioning and
answering method
¨ see the corresponding method embodiment for its specific realization
process, while no
repetition will be made in this context.
[0154] As understandable by persons ordinarily skilled in the art, realization
of the entire or
partial steps of the aforementioned embodiments can be completed by hardware,
or by a
program instructing relevant hardware, the program can be stored in a computer-
readable
storage medium, and the storage medium can be a read-only memory, a magnetic
disk, or
an optical disk, etc.
[0155] What the above describes is merely directed to preferred embodiments of
the present
invention, and is not meant to restrict the present invention. Any amendment,
equivalent
replacement or improvement makeable within the spirit and principle of the
present
invention shall all be covered by the protection scope of the present
invention.
27
Date Recue/Date Received 2022-03-22
