Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
W~ ~1/277g3 CA 02387039 2002-04-08 PCT/US00/28514
SYSTEM AND METHODS FOR ACCESSING INTERNET INFORMATION
USING INTERNET APPLIANCES
Field Of The Invention
This invention relates generally to a system
and methods for providing access to Internet
information from Internet appliances. More
specifically, the present invention provides a system
and methods for accessing summary or pre-selected items
of Internet from Internet appliances according to
users' preferences.
Background Of The Invention
The Internet and the World Wide Web
(hereinafter "the web") have revolutionized the ways in
which information is disseminated and shared. A wide
variety of information can be simultaneously accessed
by multiple users through a new category of documents
designed to easily represent content for display and
transmission over the Internet. These new documents,
often referred to as electronic documents or web pages,
are increasingly replacing their traditional paper
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-2-
counterparts as the medium through which business is
carried out.
A web page is a multimedia composition that
is displayed to the user on a "web browser window" by
"web browser software". Under the control of a user,
the web browser software establishes a connection over
the Internet between the user's computer, and a "web
server". This connection is used to download data
representing a "web page" from the web server to the
user's computer. Web pages may contain text, audio,
graphics, imagery, and video content, as well as nearly
any other type of content that may be experienced
through use of a computer or other electronic devices.
Additionally, web pages may be interactive, and may
contain user selectable links that cause other web
pages to be displayed, forms that may be used to send
information from the user to the web server,
interactive executable code, or other elements through
which the user may interact with web pages. A group of
one or more interconnected and closely related web
pages, such as all the web pages containing information
about a single company, located on one or more web
servers, is referred to as a "web site".
At present, information displayed on web
pages in the Internet can be accessed by users from
various "Internet appliances", which are electronic
devices configured with Internet access systems.
Internet appliances include, but are not limited to,
microprocessor based devices such as personal and
portable computers, personal digital assistants,
electronic organizers and toys, as well as consumer
media delivery devices such as land line and wireless
mobile telephones, television, and radio. The Internet
access systems allow users to access information such
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-3-
as their e-mail, news, stock quotes, among others, from
different Internet appliances. The information is
accessed through an interface, which can be either a
visual interface or a voice interface, depending on the
Internet appliance being used to access the
information.
A visual interface may be used in Internet
appliances that have a visual display, such as personal
and portable computers, personal digital assistants
(PDA), electronic organizers and toys, as well as
consumer media delivery devices such as wireless mobile
telephones and television. In appliances that have a
small visual display such as PDAs, electronic
organizers, and wireless telephones, access to Internet
information is provided via microbrowser software,
which consists of a simpler version of a web browser,
with reduced graphic capabilities. Users select an
information content provider in the Internet from a
pull-down menu displayed in the microbrowser window in
the wireless telephone. Such content providers include
Yahoo! from Santa Clara, CA, Excite from Redwood City,
CA, America Online, Inc., from Dulles, VA, as well
electronic commerce companies such as Amazon.com from
Seattle, WA. The user makes a specific information
request to the content provider, and the requested
information is then delivered to the user by the
microbrowser in the telephone display.
A voice interface may be used in Internet
appliances with speech recognition technologies, such
as land line and wireless mobile telephones, to access
an interactive voice response (IVR) system. These
systems enable a telephone user to access information
by dialing a phone number corresponding to an IVR
system server. The server contains software to issue
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-4-
voice prompts to the user corresponding to several
information access options, and the user responds to
the voice prompts by either selecting and pushing
buttons on the telephone, or by using speech
recognition technologies. Users can access IVR systems
from any standard telephone, including those that do
not have visual displays.
Examples of IVR systems include those
employed by financial institutions to give users access
to their accounts over the telephone, such as the
system covered in U.S. Patent Number 5,825,856.
Similar systems are also employed in a host of other
applications, including those in the airline industry
offering users the ability to check flight information
over the telephone, and the system developed by Tellme
Networks, Inc., of Mountain View, CA, that allows users
to access information on a variety of topics, such as
weather, traffic reports, restaurant reviews, stock
quotes, among others.
Both visual and voice interface systems
utilize a request-wait-response mechanism to access
information that may be interesting to a user from an
Internet appliance. The user makes requests on a very
specific topic, waits for the information to be
collected by the system from a server, and receives the
information in the visual display of the Internet
appliance in case a visual interface is used, or in the
form of voice prompts if a voice interface is used.
The request-wait-response mechanism works well on high
network and display bandwidth Internet appliances such
as personal computers, where users can quickly sift
through information and make further requests to find
exactly the information they are looking for. However,
the result is less satisfactory on low network and
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-5-
display bandwidth Internet appliances, where the speed
of information delivery is slower, presentation
capabilities are less effective, and control machinery
to interact and search for more relevant information is
difficult to use.
As an example, consider the case of users
wishing to examine their stock portfolio through a
wireless telephone. Using existing systems such as the
visual-based system designed for wireless telephones
provided by Sprint Co. from Westwood, KS, and the
voice-based system provided by Tellme Networks, Inc.,
of Mountain View, CA, the users may either individually
request details on each company in the stock portfolio
by using the telephone's limited visual or voice
interface, or the users may request a list of all the
stock information as a single stock portfolio. In both
cases, the systems do not provide access to summary or
pre-selected items of information quickly and easily.
Different pieces of information are often relevant to
different users, and users may have to iterate through
several items of less important and less interesting
information before accessing desired information.
The importance of the information may be
dependent on a combination of complex parameters, for
example, it may be determined by a combination of how
much money users have in the stock in question, what is
the percentage change in the stock price, what is the
day's range, what is the volume, and when the quarterly
results for the company issuing the stock are
announced. Users must sift through this information
using the limited interface in the telephone, which is
a very time consuming and laborious process. Further,
the telephone and Internet connections may be
unreliable, and the user may have to wait a significant
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-6-
amount of time to receive the requested information.
In view of the foregoing drawbacks of
accessing Internet information from Internet
appliances, it would be desirable to provide a system
and methods for accessing summary and pre-selected
items of Internet information quickly and easily from
Internet appliances according to users' preferences.
It further would be desirable to provide a
system and methods for creating an Internet information
database to store Internet information in categories,
manage the relative importance of different items of
information, and manage the relationships between
different items of information.
It still further would be desirable to
provide a system and methods for accessing desired
information that is relevant to a particular user's
interests at any given moment in time quickly within
the Internet information database.
It also would be desirable to provide a
system and methods for creating users' profiles to
manage the information interests of each user in
different information categories.
It further would be desirable to provide a
system and methods for delivering summary and pre-
selected items of information desired by a user in a
concise and easy to comprehend format optimized for the
network bandwidth and display capabilities of each
individual Internet appliance.
Summary Of The Invention
In view of the foregoing, it is an object of
the present invention to provide a system and methods
for accessing summary and pre-selected items of
Internet information quickly and easily from Internet
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
appliances according to users' preferences.
It is another object of the present invention
to provide a system and methods for creating an
Internet information database to store Internet
information in categories, manage the relative
importance of different items of information, and
manage the relationships between different items of
information.
It is a further object of the present
invention to provide a system and methods for accessing
the desired information that is relevant to a
particular user's interests at any given moment in time
quickly within the Internet information database.
It is also an object of the present invention
to provide a system and methods for creating users'
profiles to manage the information interests of each
user in different information categories.
It is a further object of the present
invention to provide a system and methods for
delivering summary and pre-selected items of
information to a user in a concise and easy to
comprehend format optimized for the network bandwidth
and display capabilities of each individual Internet
appliance.
These and other objects of the present
invention are accomplished by providing a system and
methods for accessing summary and pre-selected items of
Internet information from Internet appliances quickly
and easily according to a user's preferences. The user
preferences are stored in a user profile that specifies
the information interests of the user in a variety of
information categories, such as weather, news, stock
quotes, among others. The information is stored in an
Internet information database designed to enable an
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
_g_
Internet appliance to access desired information for a
given user's interests as well as other desired related
information. Information determined to be desired by a
user is delivered in a format optimized for network
bandwidth and display capabilities of that user's
Internet appliance. Such a system and methods may
automatically provide more detail on information that
is deemed relevant to an user, while less important and
less interesting information is provided only in a
summary format .
In a preferred embodiment, the system of the
present invention for providing access to pre-selected
items of Internet information from Internet appliances
according to users' preferences involve four
components: (1) an information server; (2) an
information database; (3) a user configuration
interface; and (4) an Internet appliance interface.
The information server executes the system
software and contains four software modules: (1) an
information retrieval module; (2) an information
analysis module; (3) an information classification
module; and (4) an information rendering module. The
server also stores users' profiles that specify the
information interests of each user in a variety of
information categories, such as weather, news, stock
quotes, among others.
The information retrieval module retrieves
information from diverse Internet resources such as web
servers, news feeds, and e-mail servers. The
information analysis module parses the retrieved
information to extract different representations at
varying levels of detail. The information
classification module, classifies the extracted
information according to its category and priority.
W~ 01/27783 CA 02387039 2002-04-08 PCT/LTS00/28514
-9-
The information rendering module then
extracts the information determined to be desired by a
user from the information database, and renders that
information at one or more levels of detail and
summarization dependent on that user's interest in the
information, the priority of the information, and the
information rendering capabilities of the user's
Internet appliance. The renderings of individual items
of information, which may be on a variety of categories
and at different levels of detail and summarization,
are combined into a summary delivered to the user in a
visual format and presentation optimized for the user's
Internet appliance. Optionally, the information may be
transformed to facilitate presentation using a voice
interface.
The information database is designed to
enable an Internet appliance to access information
according to a given user's interests and as well as
other related information. The database records the
information in categories and provides mechanisms to
determine the relationships between information in
different categories. These relationships define sub-
categories of information and clusters of information
that group related categories and sub-categories
together. The clusters of information may be defined
by manually programmed relationships between
information categories and sub-categories, as well as
by relationships formed when natural language
processing of the information database can determine
what information categories and sub-categories
associated with the information accessed by a user are
available.
The user specifies which information he or
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-10-
she considers relevant through a user configuration
interface. The interface enables the user to
explicitly specify not only the categories of
information of interest, but also the relative level of
interest in each information category. The information
is accessed through an Internet appliance interface in
a concise and easy to comprehend format optimized for
network bandwidth and display capabilities of each
individual Internet appliance.
Advantageously, the present invention enables
a user to access Internet information quickly and
easily from any Internet appliance.
In addition, the present invention enables a
user to select the information accessible through the
Internet appliance according to his or her personal
preferences.
Brief Description Of The Drawings
The foregoing and other objects of the
present invention will be apparent upon consideration
of the following detailed description, taken in
conjunction with the accompanying drawings, in which
like reference characters refer to like parts
throughout, and in which:
FIG. 1 is a schematic view of the system and
the network environment in which the present invention
operates;
FIG. 2 is a flow chart of the methods
employed by the information server of the present
invention;
FIG. 3 is a schematic view of the
transactional queuing mechanisms for retrieving
Internet information;
FIG. 4 is a schematic view of an illustrative
VV~ 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-11-
information query;
FIG. 5 shows a logic equation to calculate
the importance and priority of a weather report;
FIG. 6 is a schematic view of an illustrative
fragment of a categorization tree for managing
relationships between information items in the
information database;
FIG. 7 is a schematic view of an illustrative
ad hoc relationship between content nodes in a
categorization tree;
FIG. 8 is a schematic view of an illustrative
user profile;
FIG. 9 shows an equation to calculate the
importance of a content or virtual content node in a
categorization tree to an user;
FIG. 10 is a schematic view of an
illustrative configured rendering definition for the
lottery information category;
FIG. 11 is a flow chart for combining
renderings of information categories in accordance with
the principles of the present invention;
FIG. 12 is a flow chart for transforming text
to facilitate presentation using a voice interface;
FIG. 13 is a schematic view of illustrative
written text to voice transformations for improving the
understandability of text for voice-based Internet
appliance interfaces;
FIG. 14 shows an example text transcript of
user access to an illustrative voice-based Internet
appliance interface; and
FIGS. 15A and 15B are schematic views of
illustrative visual-based interfaces for use with
Internet appliances having a visual display.
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-12-
Detailed Description Of The Invention
The present invention provides a system and
methods for accessing concise, relevant, and important
items of Internet information quickly and easily from
Internet appliances according to an user's preferences.
Internet appliances, as used herein, are electronic
devices configured with Internet access systems.
Internet appliances include, but are not limited to,
microprocessor based devices such as personal and
portable computers, personal digital assistants,
electronic organizers and toys, as well as consumer
media delivery devices such as land line and wireless
mobile telephones, television, and radio.
Referring to FIG. 1, a schematic view of the
system and the network environment in which the present
invention operates is described. Internet appliances
(personal computer), 21 (portable computer), 22
(mobile telephone), 23 (land-line telephone), 24 (PDA),
and 25 (television) communicate across the Internet
20 with information server 26. Internet appliances 20-25
have an Internet appliance interface to access Internet
information in a concise and easy to comprehend format
optimized for the network bandwidth and display
capabilities of each individual appliance.
The Internet appliance interface may be a
visual interface or a voice interface, depending on the
Internet appliance being used to access the
information. A visual interface may be used in
Internet appliances that have a visual display, such as
personal computer 20 and portable computer 21, wireless
mobile telephone 22, PDA 24, and television 25. A
voice interface may be used in Internet appliances with
speech recognition technologies, such as wireless
mobile telephone 22 and land line telephone 23.
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-13-
The Internet appliance interfaces access
Internet information by connecting to information
server 26, which executes the system software and is
responsible for retrieving information from diverse
Internet resources such as web servers, news feeds, and
e-mail servers. Information server 26 parses the
retrieved information to extract different
representations at varying levels of detail and
classify the extracted representations according to
their category and priority.
Information server 26 also stores users'
profiles that specify the information interests of each
user in a variety of information categories, such as
weather, news, stock quotes, among others, and extracts
the information that is currently most important to the
user from information database 27. Information server
26 combines the extracted information into a concise
and easy to understand summary to be delivered to the
user's Internet appliance.
Information database 27 enables Internet
appliances 20-25 to access the most important
information for a given user's interests. Information
database 27 stores the information retrieved by
information server 26 in categories and provides
mechanisms to determine the relationships between
information in different categories.
Users of Internet appliances 20-25 specify
which information they consider to be important and
relevant to them through user configuration interface
28. User configuration interface 28 enables users to
explicitly specify not only the categories of
information in which they are interested, but also
their relative level of interest in each information
category. User configuration interface 28 may be a
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-14-
user configuration web site, where a questionnaire
about the user's interests can be completed.
Information collected in the questionnaire is
transmitted to information server 26, which translates
the collected information into a user profile. User
configuration interface 28 also may be a call center
where a user talks to a human operator who asks
questions about the user to build up the user profile.
It will be apparent to one of ordinary skill
in the art that Internet appliances 20-25 are
illustrative, and that alternative Internet appliances
may be used with the present invention. It will
further be understood that the present invention also
could be used in other network settings. For example,
rather than connecting through the Internet, the system
and methods of the present invention could be used on a
local area network. In such a configuration, the
Internet appliances and server would all be connected
to the same local area network.
I. Overview of Information Flow from the Internet to an
Internet Appliance
Referring to FIG. 2, a flow chart of the
methods employed by the information server to deliver
concise, relevant, and important information from the
Internet to an Internet appliance is described. At
step 30, the information retrieval module of the
information server reliably retrieves information from
diverse Internet resources such as web servers, news
feeds, and e-mail servers. The information may contain
news and weather reports, stock quotes, e-mails, among
others. The retrieved information is parsed at step 31
by the information analysis module of the information
server to extract different representations of
W~ ~l/277g3 CA 02387039 2002-04-08 PCT/US00/28514
-15-
information at varying levels of detail. For example,
a stock quote for a company may be extracted solely by
the stock price at a low detail representation, or by
the stock price, the low and high of the day's price,
the volume traded, and the percent change in value at a
higher detail representation.
At step 32, the extracted information is
classified by the information classification module of
the information server according to the category and
priority of the information. The information category
may include weather, traffic, financial news, movies,
among others. The priority of the extracted
information is based on the importance of the
information independent of the level of interest any
individual user has in that information and on the
importance of the information independent of how old
the information is. The priority is calculated based
on the assumption that the information has just been
reported.
At step 33, the information is stored in the
information database. The information database is
designed to enable an Internet appliance to access the
most important information for a given user's interests
as well as other important inter-related information.
The database records the information in categories and
provides mechanisms to determine the relationships
between information in different categories. These
relationships define sub-categories of information and
clusters of information that group related categories
and sub-categories together. The clusters of
information may be defined by manually programmed
relationships between information categories and sub-
categories, as well as by relationships formed when
natural language processing of the information database
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-16-
can determine what information categories and sub-
categories associated with the information accessed by
a user would ideally be available.
Lastly, at step 34, the information rendering
module extracts from the information database the
information determined to be important to the user.
The information extracted from the database is rendered
at different levels of detail and summarization
dependent on the user's interest in the information,
the priority of the information, and the information
rendering capabilities of the user's Internet
appliance. The user's information preferences are
extracted from the user's profile stored in the server.
The user's profile specifies the information interests
of each user in a variety of information categories,
such as weather, news, stock quotes, among others. The
renderings of individual items of information, which
may be on a variety of categories and at different
levels of detail and summarization, are combined into a
concise and easy to understand summary delivered to the
user at step 35 in a format and presentation optimized
for the user's Internet appliance.
II. Information Retrieval
The information retrieval module of the
information server reliably retrieves information from
diverse Internet resources such as web servers, news
feeds, and e-mail servers. To ultimately provide a
robust and deterministic user interface for the user,
availability of up to date Internet information must be
guaranteed by the system. Standard Internet protocols
do not guarantee end to end delivery of information
since any number of computer or network failures can
effect the delivery. Unless the user is protected from
W~ ~l/27~g3 CA 02387039 2002-04-08 PCT/US00/28514
-17-
this unreliability, speed and ease of use will be
compromised. The information retrieval module uses
transactional queuing mechanisms with retry queues to
manage requests for Internet information and handle
failures. This ensures that required information
content is brought into the system with the appropriate
or necessary frequency to satisfy user demand.
Referring to FIG. 3, a schematic view of the
transactional queuing mechanisms for retrieving
Internet information is described. The information to
be retrieved from the Internet by the information
server is specified in a configuration object called an
information query.
An information query object specifies the
location on the Internet from which the information may
be fetched, for example, the URL of a web site, the
frequency during which it should be collected, for
example, every hour during the day, the category of the
information being collected, how to calculate the
priority of the information being collected, and how
the retrieved information should be parsed by the
information analysis module of the information server,
that is, what specific information items should be
extracted from the information collected.
An information query can specify an
individual item of Internet information to be
collected, or more often, it can specify a collection
of Internet information such as a set of stock prices
from a user's portfolio or even all the stock prices on
the New York Stock Exchange.
Two queues are used to manage the information
requests from the Internet, normal queue 36 and retry
queue 37. Normal queue 36 handles the successful
information retrievals, while retry queue 37 handles
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-18-
the failed information retrieval attempts. Information
queries are retrieved from the Internet starting at the
head of normal queue 36 and finishing at the tail of
normal queue 36. The same information query may be
placed at different positions in normal queue 36 to
retrieve the information at a particular time
frequency.
A request for information query 38 at the
head of normal queue 36 is sent to the Internet by the
information server. There are two possible outcomes of
this request. First, the information specified in
information query 38 is retrieved from the Internet
within a configured allowed time and so processing of
the information by the information analysis and
information classification modules of the information
server and storage of the information in the
information database can be completed.
Upon completion, information query 38 is
placed at the tail of normal queue 36 as information
query 39. Second, either a network failure occurs
during retrieval of information query 38 or the
information is not retrieved from the Internet within a
configured time limit. In this case, information query
38 is placed at the tail of retry queue 37 as
information query 40 to be retried at a later date.
Retry queue 37 is processed in a similar way
to normal queue 36, with a request for information
query 41 at the head of retry queue 37 being sent to
the Internet. Information query 41 consists of a
failed information request that was previously
attempted by the information server in normal queue 36.
Similarly to normal queue 36, there are two possible
outcomes for requesting information query 41.
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-19-
First, the information specified in
information query 41 is retrieved from the Internet
within a configured allowed time and so processing of
the information by the information analysis and
information classification modules of the information
server and storage of the information in the
information database can be completed. Upon
completion, information query 41 is placed at the tail
of normal queue 36 as information query 39. Second,
either a network failure occurs during retrieval of
information query 41 or the information is not
retrieved from the Internet within a configured time
limit. In this case, information query 41 is placed at
the tail of retry queue 37 as information query 40 to
be retried at a later date.
In such a way, normal queue 36 handles
collection of information that can be handled
"reliably," i.e., where information can be retrieved
with a statistically acceptable frequency, and retry
queue 37 handles "unreliable" Internet information.
The information server uses this basic queuing
mechanism to ensure transactional receipt of Internet
content and efficient handling of network and Internet
server failures. Advantageously, failure to retrieve
information from the Internet is dealt with gracefully,
that is, without data loss by the scheduling of a retry
in retry queue 37.
In addition, careful configuration of time
out periods on the processing of information queries
allows efficient information retrievals. For example,
a relatively short time out period on normal queue 36
means that slow and unreliable connections are aborted
quickly, leaving normal queue 36 to continue to rapidly
process other information queries. A long time out
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-20-
period on retry queue 37 means that the information
server will wait for slower connections without holding
up the receipt of the main bulk of Internet
information. Further efficiency gains may be made
through parallelism, both in the use of different queue
pairs for different types of information content, for
example, one queue pair to handle e-mails, another
queue pair to handle web sites, and so on, and also by
having multiple mufti-threaded processors of each
queue.
III. Information Analysis
Once information has been received from the
Internet, it is parsed by the information analysis
module of the information server to extract different
representations of information at varying levels of
detail. The different representations of information
to be extracted are specified in the information query
issued by the information retrieval module of the
information server.
Referring now to FIG. 4, a schematic view o~
an illustrative information query is described.
Information query 42 consists of an XML function that
is executed by the information server to extract stock
quote information from the Internet. Information query
42 contains two fields, query field 43 and information
field 44. Query field 43 specifies URL 45 from which
the stock price information is to be extracted, broad
category 46 of the information being collected, time 47
during which the information should be collected, and
how the priority of the information should be
calculated (48). Upon execution of information query
42, the information server receives the source data
contained in URL 45.
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-21-
Information field 44 specifies how to parse
the collected information, that is, what data fields
(49) should be extracted from retrieved URL 45 and how.
Data extraction utilizes pattern-matching techniques to
extract data fields 49, which in the case of
information query 42 for a stock price, include: the
name of the stock, the market in which the stock is
traded, the stock symbol, the last trade date and
price, the change and change percent in value of the
stock, the previous closing price, the volume traded on
the day, among others. The process for extracting data
fields 49 involves searching for known terms and
markers in URL 45. These terms and markers are used as
beginning and end delimiters for the data field being
sought. For example, data field 50 for the stock
symbol specifies data type 51, beginning delimiter 52,
and end delimiter 53.
The pattern matching specification can either
be built manually by a system administrator operating a
user interface to highlight data fields 49 in URL 45,
or automatically by using natural language analysis and
heuristic techniques to recognize expected data fields
in a source information of a known context. In case
the information is easily accessed from a database, the
natural language analysis is not required since the
data fields can be directly extracted from the
database. The natural language analysis is preferred
to the manual method since it does not involve human
interaction, which would be very slow and inefficient
in retrieving Internet information during short time
intervals.
WO 01/27783 CA 02387039 2002-04-08 pCT/US00/28514
-22-
IV. Information Classification
The information classification module
classifies the extracted information according to its
category and priority. Classifying the information
according to its category begins with knowledge of
where the information collected originated. In some
circumstances, this may be all that is required to
position the information in the correct information
category. For example, if the information came from a
news feed that provides only surf reports for a
specific beach, then the information can be
automatically categorized.
More likely, knowledge of the source
information provides only broad categorization, and the
final classification can only be determined in
combination with data fields extracted during the
information analysis. For example, a news feed
providing stock quotes specifies a broad classification
of any information read from the news feed as a stock
quote, but this broad categorization must be combined
with the company name and perhaps the market on which
the stock is traded to provide the final category
classification. This additional information can be
determined from the data fields extracted during the
information analysis, with the rules for providing the
final classification category being specified in the
information query. For example, information query 42
is categorized based on a combination of a broad
category field and the name of the stock (54).
The information classification module
calculates the priority of each item of Internet
information collected based on the importance of the
information independent of the level of interest any
individual user has in that information and on the
WO 01/27783 CA 02387039 2002-04-08 pCT/LTS00/28514
-23-
importance of the information independent of how old
the information is. The priority is calculated based
on the assumption that the information has just been
reported. An individual user's interest in a category
of information and the modification of priority based
on how recently the information was reported are
properties that are applied by the information
classification module at the time at which the
information is accessed by the user. The individual
user's interest in a category of information is
specified in the user's profile stored in the
information server.
The priority of an information category is
computed using arithmetic and logic equations. For
example, illustrative priority equation 48 in FIG. 4 is
an arithmetic equation to calculate the importance of a
stock quote based on the percentage change of a stock
price. Data used within the calculation is taken from
the CHANGEPERCENT data field (49) extracted during
information analysis.
Referring now to FIG. 5, an illustrative
logic equation to compute the importance and priority
of a weather report is described. Logic equation 55
determines the importance and priority of a weather
report by looking for occurrences of unusual weather
conditions within the weather report, such as
thunderstorms, ice, sleet, wind conditions, among
others. When an unusual weather pattern is present at
the weather report, the priority of the weather report
is increased.
The priority of an information category is
represented by a numeric value such that the higher the
value, the higher the priority. The priority can take
values between 0 o and 100 0, and it is normalized to
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-24-
provide consistent values across information
categories. An assignment of a priority of 95 0
implies that an item of information is more important
than 95 0 of the information previously collected in a
specific information category. In this way, historical
trends can be used to normalize the calculation of
priority of an item of information within a category.
It should be understood by those skilled in
the art that a wide variety of equations may be used to
calculate the priority of any given information
category. For example, in calculating the priority of
e-mail received, header information within the e-mail
can be used as the basis of a calculation of the
priority of an individual message. Such information
can be extracted from the subject of the e-mail by
looking for reply or forward markers, or from the e-
mail distribution list by looking for e-mail sent
directly to the user.
V. Information Database
Once the retrieved information has been
analyzed and classified according to its category and
priority, it is stored in an information database
designed to enable an Internet appliance to access the
most important information for a given user's interests
as well as other important inter-related information.
The information database records the information in
categories and provides mechanisms to determine the
relationships between information in different
categories.
Referring to FIG. 6, a schematic view of an
illustrative fragment of a categorization tree for
managing the relationships between information items in
the information database is described. Categorization
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-25-
tree 56 organizes the information in the information
database by categories and associated relationships
between information items in the categories. The
categories and associated relationships are manually
programmed in the information database.
Each information item in categorization tree
56 is a node in the tree, and a universally unique
identifier is assigned to every node in the tree to
uniquely identify a node. There are five types of nodes
in categorization tree 56: (1) global root node; (2)
category node; (3) cluster definition node; (4) content
node; and (5) virtual content node. Each node in the
tree may be of more than one type. Global root node 57
defines the head of the information space from which
all of the information in the database may be
traversed. Category nodes 59a and 59b represent a
category of information within the information
database, with category node 59a representing the
entertainment category and category node 59b
representing the geography category. Category node 59a
is divided into category nodes 61a-c, which are sub-
categories of the entertainment category, namely,
movies (61a), producers (61b), and actors (61c).
Category node 61a refers to content node 63a
and virtual content nodes 63b and 65. Content node 63a
contains an instance of information content produced by
the information analysis module of the information
server, in this case, a video clip of a movie. Content
nodes may be of different types, including text,
graphics, imagery, video, sound, and combinations.
Virtual content nodes 63b and 65 specify instances of
information that vary depending on context parameters,
for example, virtual content node 63b represent movie
theater information that depends on location, and
W~ 01/27783 CA 02387039 2002-04-08 PCT/L1S00/28514
-26-
virtual content node 65 represent movie schedules that
depend on date and time. Other virtual content nodes
in categorization tree 56 include node 61d representing
maps of given locations, and node 61e, representing
driving directions between two locations.
In addition, category nodes 61a (movies)and
virtual content node 63b (theaters) are also cluster
nodes, which define a cluster of information containing
related information items. Cluster node 61a define
cluster 66 and cluster node 63b define cluster 67.
Cluster 66 is used to specify that when information on
a movie is requested by a user, related information
such as the movies' producers (61b) and actors (61c), a
sample preview clip of the movie (63a), and the
theaters where the movie is currently playing (63b) can
also be displayed to the user. Information clusters
allow the users to see all relevant information on a
topic and can be cascaded to form large interrelated
maps of content. When a new information item belonging
to an information category is inserted in the
information database, any existing information cluster
for that information category is checked to determine
whether the new information item can be inserted in the
cluster.
The nodes in categorization tree 56 are
connected to each other through several arcs that
define relationships between the nodes. An arc links
two nodes specified by their unique identifiers, and
also has properties defining the weighted strength of a
relationship, the age and any expiration date of the
relationship, and any ownership or privacy rules. The
weighted strength of a relationship is a value
associated with the arc that determines how strongly
the nodes connected by the arc are related. This value
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-27-
is used by the information rendering module of the
information server when extracting the most important
information for a user from the information database.
Ownership and privacy properties define which users
have access to the information stored at the node
destination of an arc. For example, e-mail messages
are made private to a user by their ownership being
associated to the user and properties being set on the
arc that enforce the privacy of e-mail messages.
There are five types of arcs in a
categorization tree: (1) "is a subcategory of" arc; (2)
"is an instance of" arc; (3) "is equivalent to" arc;
(4) "is related to"; and (5) "is a cluster definition
of" arc. The "is a subcategory" arc is used to link
the global root node to category nodes to form a
categorization of the information space. Arcs 58a and
58b in categorization tree 56 are of this type.
The "is an instance of" arc is used to
connect category nodes to content nodes or virtual
content nodes to indicate that the content is in the
specified category. Arcs 62a and 62b in categorization
tree 56 are of this type.
The "is equivalent to" arc indicates when two
or more content nodes contain equivalent content, for
example, two content nodes containing news stories on
the same topic and same date but from different news
sources would be connected with an "is equivalent to"
arc.
The "is related to" arc is used to specify
relationships between instances of information in
content nodes and virtual content nodes, such as arc 64
in categorization tree 56.
Lastly, the "is a cluster definition of" arc
is used to connect a category node to a cluster node
WO 01/27783 CA 02387039 2002-04-08 PCT/LTS00/28514
-28-
and that cluster node to other nodes that may contain
information that if available should be presented with
the requested information in the cluster node since it
is closely related. Arcs 68a-d in cluster 66 of
categorization tree 56 and arcs 69a-c in cluster 67 of
categorization tree 56 are of this type.
Information items in the information database
also can be interrelated by defining "ad hoc
relationships" between them. Ad hoc relationships are
formed when natural language processing of content
nodes in a categorization tree can determine what other
associated information are available to an user and
also what information a content node can supply that
would be of interest to other content nodes. This
enables ad hoc relationships to be built between
content nodes from anywhere within the information
database without having to follow manually programmed
cluster or category definitions already in place in the
information database.
Referring now to FIG. 7, a schematic view of
an illustrative ad hoc relationship between content
nodes in a categorization tree is described. Content
node 70 contains an instance of a summary plot of the
movie Antz, and is connected to content nodes 71, 72,
73, and 74, through "is related to" arcs. Natural
language processing of the information in the
categorization tree has determined that content node 75
representing a TV show containing a review of the movie
Antz is related to content node 70. An ad hoc
relationship is then formed between content nodes 70
and 75 though "is related to" arc 77.
W~ ~l/277g3 CA 02387039 2002-04-08 PCT/US00/28514
-2 9-
VI. User Configuration Interface & User Profile
A user specifies which information he or she
considers to be important and relevant through a user
configuration interface. The interface enables users
to explicitly specify not only the categories of
information of interest, but also the relative level of
interest in each information category. The user
configuration interface may be a user configuration web
site or a call center where a user talks to a human
operator. In both cases, a questionnaire about the
user's interests is completed and transmitted to the
information server, which translates the collected
information into a user profile.
Referring to FIG. 8, a schematic view of
an illustrative user profile is described. User
profile 78 consists of an XML function that is executed
by the information server when the user requests
information from an Internet appliance. The user
profile is used to extract only the information
determined to be important and relevant to that user
from the information database.
User profile 78 specifies user's interests in
categories 79a-f, namely, e-mail (79a), financial
(79b), lottery (79c), traffic (79d), weather (79e), and
horoscope (79f). Category 79b has two sub-categories,
80a and 80b, further specifying the user's interest in
stock prices of technology companies as the relevant
financial information. Categories 79a-f have
information or content items 81a-h that specify exactly
what information item belonging to a given category the
user has an interest in. For example, in traffic
category 79d, the user only has an interest in the
traffic of the San Francisco Peninsula in California,
USA (81f).
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-30-
Information items 81a-h have associated
interest levels 82a-h that specify the importance and
priority of information items 81a-h to the user.
Interest levels 82a-h can be adjusted dynamically based
on the user's behavior while interacting with an
Internet appliance interface. For example, if
information is presented to a user on a particular
category of information and the user routinely skips
past that information in the Internet appliance
interface, the system may determine that the user has a
reduced interest in this category of information. As
such, the interest level on that particular category
recorded in the user profile may be reduced accordingly
with appropriate notification to the user.
Conversely, if a user spends a significant
amount of time accessing a particular category of
information in the Internet appliance interface, it can
be reasonably determined that the user has a higher
interest level in that information category. As such,
the interest level on that particular category recorded
in the user's profile can be increased accordingly with
appropriate notification to the user. The dynamic
adjustments to a user's interest levels are limited
within reasonable bounds to avoid such adjustments
having a more significant effect on the user profile
than more accurate specifications of a user's interests
specified in questionnaires through user configuration
interfaces.
VII. Information Rendering
The user's profile is combined with the
information in the information database to provide the
most important information to the user through an
Internet appliance interface. The information
rendering module of the information server traverses an
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-31-
individual user's profile, taking each definition of a
category of interest as a starting point within the
information database, which corresponds to a category
node in a categorization tree. From each of these
starting points within the information database, a
calculation is carried out to determine which pieces of
information in the vicinity of the starting point are
important enough to bring to the user's attention, that
is, which content nodes and virtual content nodes can
be reached from the starting point by traversing arcs
in the categorization tree in the information database.
The importance of a content node and virtual content
node to the user and how strongly the nodes are related
depend on the combined weighted strength of the arcs
that are traversed.
Referring to FIG. 9, an illustrative equation
to compute the importance of a content or virtual
content node in a categorization tree to an user is
described. Equation 83 involves three parameters, 83a-
c, that are used to calculate the importance of a
content or virtual content node. Equation parameter
83a consists of the multiple of all strength of
relationship weights assigned to the arcs that are
traversed from the starting point to get to the content
or virtual node whose importance is being determined.
The weights are recorded in the information database.
Equation parameter 83b is extracted from the user's
profile and consists of the interest level of the user
in the information category represented by the category
node at the starting point. Lastly, equation parameter
83c is an aging factor stored in the information
database for each arc traversed from the starting point
to the content or virtual node whose importance is
being determined. The aging factor for a given arc is
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-32-
determined as the difference in dates between the date
when the arc was created and the date when the user is
accessing the information at the starting point.
Equation 83 is calculated for each of the
user's interests so that a prioritized list of the
importance of the information in the information
database to be transmitted to the user is constructed.
The prioritized list is used to determine the order in
which the information is to be transmitted to the user
when accessing an Internet appliance interface. The
process of constructing the prioritized list can be
optimized by using importance thresholds for arcs that
are traversed from a starting point to a content or
virtual content node. Arcs whose importance fall below
the importance threshold will link to information items
that will not be important enough to bring to the
attention of the user, so that equation 83 is only
calculated for the subset of arcs which is most
interesting to the user.
By adjusting the importance threshold, the
quantity of information pulled from the information
database can be adjusted for different Internet
appliance bandwidths, different Internet appliance
interfaces, and different user requests. Additionally,
when this process is applied to only a subset of the
user's profile, a similar prioritized list can be
constructed for a subset of information categories in
the user's profile. These subsets allow the user to
have more control of the level of detail presented for
each information category when accessing an Internet
appliance interface. For example, when a user
initially logs into an Internet appliance interface a
high level summary can be obtained of all their
information interests at a relatively high importance
WO U1/2778~ CA 02387039 2002-04-08 PCT/US00/28514
-33-
threshold. The user then may go on to view more
information, including that at a lower importance
threshold in a specific information category, such as
viewing the details on a stock portfolio when accessing
the financial category.
The calculation of the importance of each
individual item of information to a user allows a
decision to be automatically made as to whether that
information item should be presented to the user when
accessing a particular Internet appliance. It also
enables the information rendering module of the
information server to decide how much detail should be
provided when the information is presented (or
"rendered") to the user such that an information item
that is more important to the user at a moment in time
is rendered with more detail than another information
item that is less important. This approach helps the
information rendering module of the information server
present information concisely to the user.
The rendering of information is based on
configured rendering definitions that are stored in the
information server and specify what information is
available at different levels of detail, or strata. A
rendering of information at a higher strata is more
summarized than a rendering of the same information at
a lower strata.
Referring to FIG. 10, a schematic view of an
illustrative configured rendering definition for the
lottery information category is described. Rendering
definition 84 consists of an XML function that is
executed by the information server when information is
to be rendered from the information database to an
Internet appliance. Rendering definition 84 contains
two rendering levels of detail, 85 and 86. Each
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-34-
rendering level of detail defines an information
category (85a and 86a), a stratum (85b and 86b), the
Internet appliance interface for which the information
is being rendered (85c and 86c), and the language in
which it is being rendered (85d and 86d).
Each rendering level provides formatting
within which data parameters such as data parameter 87
are substituted. Data parameters substituted in
renderings are extracted by the information analysis
module of the information server and stored in the
information database. Rendering definitions may be
conditional on the values of data parameters such that
detail for a particular information category is only
rendered if the values of certain data parameters
indicate that the user may be interested in receiving
the detailed information. For example, detail on the
volume of a stock may be rendered to an user if a stock
volume parameter indicates that the volume in the
traded stock has reached an exceptional value at a
particular date.
Stratum 85b at 100 o represents a minimum
rendering consisting of only a summary of the latest
information on information category 85a. Stratum 86b
at 50 % provides some additional detail for the user,
in this case, the latest two lottery results, that may
be rendered if the user's interest level in information
category 86a is high enough. Any number of renderings
may be configured between the 100 % stratum and 0%
stratum levels. A rendering for an item of information
is selected by the information rendering module of the
information server when the stratum value is
appropriate for the level of importance of the
information to the user. The higher the importance of
an item of information, the lower the stratum and the
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-35-
lower the summarization selected to render the
information, that is, the information will be rendered
with more detail.
The selection of a rendering definition
depends on the Internet appliance interface (85c and
86d) for which information is being rendered. Only
renderings appropriate for a user's Internet appliance
are selected for rendering an item of information. The
same information item may be rendered in different ways
for different Internet appliances so as to optimize the
effectiveness of the information on the target
appliance. For example, a stock quote may be rendered
in the form of a visual graph on an Internet appliance
with a visual display, but on a voice-based Internet
appliance the rendering may consist of only the
percentage change in the stock price.
Referring to FIG. 11, a flow chart for
combining renderings on a variety of information
categories into a concise and easy to understand
summary to be delivered to an Internet appliance is
described. To make the combined rendering both easy to
understand by the user and also most effective in
quickly providing the most important information to an
user, a two-stage sorting algorithm is carried out.
First, at step 89, the list of renderings of all
information categories that are important to the user
are sorted so that the most important information
categories are rendered to the user before the less
important ones. Second, at step 90, a sort is
performed within each information category to order the
individual information items in each information
category by their importance to a user. For example,
ordering individual information items in the stock
quotes category would render a stock price that has
WO ~l/27783 CA 02387039 2002-04-08 PCT/US00/28514
-36-
risen substantially prior to one that is stable.
Providing a concise and easy to understand
summary optimized for network bandwidth and display
capabilities of each individual Internet appliance
optionally involves a method for transforming text so
that it can be presented using a voice-based interface.
Referring now to FIG. 12, a flow chart for
transforming written text so that it may be presented
by a voice interface as it is normally spoken in
conversation is described. At step 93, an information
context analysis is performed so that the broad context
of the information is determined by the information
rendering module. The broad context may determine, for
example, that the information item is an e-mail
message, or that the information item contains
telephone numbers of a movie theater.
The context analysis searches for known
keywords within the information rendering that indicate
the context of the information, for example, a
paragraph of information that contains the words
"highway" and "interstate" is likely to be discussing
roads, a sentence that starts with the word "phone" is
likely to contain a phone number, a sentence that
contains the word "dollar" or the character "$" is
likely to contain financial figures, and so on. At
step 94, a word token analysis is performed by the
information rendering module to determine word tokens
whose understandability when spoken in conversation can
typically be improved. Examples of such word tokens
include streams of numbers or dates, or header
information in an e-mail message.
By combining steps 93 and 94, a high degree
of reliability can be achieved in improving the
understandability of text. Referring now to FIG. 13, a
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-37-
schematic view of illustrative written text to voice
transformations for voice-based Internet appliance
interfaces is described. Information context analysis
96, token analysis 97, and improved renderings 98
contain examples of transformations of different
information items in particular contexts, including
information items in financial context 99 and
information items in telephone context 100.
Information items in financial context 99 are
rendered in such a way to clarify specific symbols used
in a financial context, such as "$" signs and "
fractions of stock quotes, which are rendered in speech
as the word "dollar" and the word "half", respectively.
Similarly, telephone numbers in telephone context 100
are rendered so that well-known phone prefixes, such as
"1-800" and area codes, are easily understood in a
voice-based interface.
VIII. Internet Appliance Interface
The rendered information can be accessed from
an Internet appliance through an Internet appliance
interface, that may be a voice interface or a visual
interface, depending on the display capabilities of the
Internet appliance being used. The Internet appliance
interface is designed to that information can be
accessed from any Internet appliance with little user
interface interaction. The Internet appliance
interface preferably provides the a high degree of
desired Internet information to an user in a
conveniently accessible manner from any Internet
appliance.
Referring to FIG. 14, an example text
transcript of user access to an illustrative voice-
based Internet appliance interface is described. Text
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-38-
transcript 101 contains the text translation of voice
prompts issued by the user of a voice-based Internet
appliance interface as well as the text translation of
the voice information sent by the information server to
the user during a user's phone call to the server.
Upon dialing the phone number corresponding to the
server, the user receives summary 102 of the
information important to the user. After a brief pause
following summary 102, the information server presents
the user menu of options 103 to prompt the user for
access to the user's pre-selected information
categories.
When the user selects an information category
option by saying the number corresponding to the
selected information category, the server fetches the
appropriate rendering configuration of the information,
and translates the information to the user through
information voice notification 104. The information
transmitted to the user for each information item
depends on the importance of the information item, with
more detail information being provided for the
information items that are considered more important.
A brief pause follows information voice
notification 104, and the user then hears menu 103
again to select further information categories. The
user selects the e-mail information category, and
listens to e-mail notification 105 from the server.
Selecting the e-mail information category again after
listening to e-mail notification 105 triggers the
server to render the important messages in e-mail
notification 106 providing more detail to the user. At
any point during the phone call the user may choose to
hangup, or to return to menu of options 103 to access
more detailed information on the relevant information
WO 01/27783 CA 02387039 2002-04-08 PCT/US00/28514
-39-
categories.
Referring to FIGS. 15A and 15B, schematic
views of illustrative visual-based interfaces for use
with an Internet appliance having a visual display is
described. Visual-based interface 107 shown in FIG.
15A contains top-level summaries on various information
categories of interest to an user named Dave, including
e-mail messages, financial news, weather information,
shopping requests, news headlines, and entertainment
news. Each information category is easily accessed
through icons displayed in visual-based interface 107.
Accessing the information category icons enables the
user to get more detail information on any one of the
relevant information categories. Visual-based
interface 108 shown in FIG. 15B consists of an instance
of user interaction requesting information on the movie
Antz. Visual-based interface 108 displays all the
related information on the movie Antz that is available
on the information database.
Although particular embodiments of the
present invention have been described above in detail,
it will be understood that this description is merely
for purposes of illustration. Specific features of the
invention are shown in some drawings and not in others,
and this is for convenience only and any feature may be
combined with another in accordance with the invention.
Steps of the described processes may be reordered or
combined, and other steps may be included. Further
variations will be apparent to one skilled in the art
in light of this disclosure and are intended to fall
within the scope of the appended claims.
