Note: Descriptions are shown in the official language in which they were submitted.
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
Method, System and Graphical User Interface for Providing
Reviews for a Product
TECHNICAL FIELD
[0001] The disclosed embodiments relate generally to search engines. More
particularly, the disclosed embodiments relate to methods, systems, and
graphical user
interfaces for finding, aggregating, and providing reviews for a product.
BACKGROUND
[0002] Many Internet users research a product (e.g., a Canon digital camera)
or a
service (e.g., a hotel room in Hawaii) before obtaining it. Currently, the
approach that many
users follow is to use Internet search engines. Users issue a search query
that contains the
product or service name plus terms like "review" and sift through the myriad
of results that
are returned by the search engine. While some of these results do contain
reviews about the
product, many results do not. Moreover, it is very difficult to get a holistic
view of all the
reviews. In short, this process is fairly cumbersome, time consuming, and
inefficient.
[0003] Alternatively, users may go to Web sites that provide reviews and
ratings for
products, such as www.amazon.com, www.epinions.com, and www.circuitcity.com. A
few
sites, such as www.rottentomatoes.com, movies.yahoo.com, and
www.consumersearch.com,
aggregate reviews for the same product from several sites. However, some of
these sites
aggregate reviews manually, which is very inefficient and limits the number of
reviews that
can be gathered for each product. Moreover, all of these sites are very
limited in the tools
they provide users to navigate through the reviews.
[0004] Thus, it would be highly desirable to provide tools that enable users
to more
efficiently conduct research on the products and services they are interested
in obtaining (e.g.,
by purchase, lease, rental, or other similar transaction).
SUMMARY
[0005] The embodiments disclosed herein include new, more efficient ways to
collect
product reviews from the Internet, aggregate reviews for the same product, and
provide an
aggregated review to end users in a searchable format.
1
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
[0006] In one aspect of the invention, a server collects information
containing product
reviews for a plurality of products and automatically extracts the product
reviews from the
collected information. For at least some of the extracted product reviews, the
server identifies
a particular product that is associated with the extracted product review. For
each particular
product in at least a subset of the plurality of products, the server
generates aggregated review
information for the particular product based on a plurality of extracted
product reviews that
are associated with the particular product. The server stores the extracted
product reviews
and the aggregated review information.
[0007] In another aspect of the invention, a server collects product reviews
for a
plurality of products and automatically identifies particular products that
are associated with
particular product reviews. For each particular product in at least a subset
of the plurality of
products, the server automatically generates aggregated review information for
the particular
product including frequently appearing phrases in the product reviews
associated with the
particular product. The server automatically stores the product reviews and
the aggregated
review information.
[0008] In another aspect of the invention, a client sends a request for an
aggregated
review for a product, receives the aggregated review, and displays the
aggregated review. The
aggregated review includes portions of reviews for the product and a search
input area for
entering search terms to search for reviews of the product that contain the
search terms.
[0009] In another aspect of the invention, a server receives a search request
from a
client computer to search within a plurality of reviews for a particular
product, obtains
information satisfying the request, and sends at least a portion of the
obtained information to
the client computer. The sent information includes at least portions of
reviews for the
particular product that satisfy the search request.
[0010] Another aspect of the invention is a graphical user interface on a
computer that
includes a plurality of portions of reviews for a product and a search input
area for entering
search terms to search for reviews of the product that contain the search
terms.
[0011] As used in the specification and claims, the word "product" refers to
both
products and services. Moreover, "product" encompasses virtually any product,
service or
combination thereof that can be bought, leased, rented, or similarly obtained.
Exemplary
products include, without limitation, consumer products, business products,
movies, books,
restaurants, hotels, and travel packages.
2
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
BRIEF DESCRIPTION OF THE DRAWINGS
[0012] For a better understanding of the aforementioned aspects of the
invention as
well as additional aspects and embodiments thereof, reference should be made
to the
Description of Embodiments below, in conjunction with the following drawings
in which like
reference numerals refer to corresponding parts throughout the figures.
[0013] Figure 1 is a block diagram illustrating an exemplary distributed
computer
system according to an embodiment of the invention.
[0014] Figure 2A is a block diagram illustrating an aggregated reviews backend
server 102 in accordance with one embodiment of the present invention.
[0015] Figure 2B is a block diagram illustrating an aggregated reviews
frontend
server 100 in accordance with one embodiment of the present invention.
[0016] Figure 3 is a flowchart representing a method for collecting and
aggregating
reviews for a product according to one embodiment.
[0017] Figure 4 is a flowchart representing a method for displaying an
aggregated
review for a product according to one embodiment.
[0018] Figure 5 is a flowchart representing a method for searching aggregated
reviews for a product according to one embodiment.
[0019] Figures 6A-6D are schematic screen shots of an exemplary graphical user
interface of a product information module (Figure 6A) and a product review
presentation
module (Figures 6B-6D) according to one embodiment.
DESCRIPTION OF EMBODIMENTS
[0020] Methods, systems, and graphical user interfaces are described that show
how
to find, aggregate, and provide reviews for a product. Reference will be made
to certain
embodiments of the invention, examples of which are illustrated in the
accompanying
drawings. While the invention will be described in conjunction with the
embodiments, it will
be understood that it is not intended to limit the invention to these
particular embodiments
alone. On the contrary, the invention is intended to cover alternatives,
modifications and
equivalents that are within the spirit and scope of the invention as defined
by the appended
claims.
3
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
[0021] Moreover, in the following description, numerous specific details are
set forth
to provide a thorough understanding of the present invention. However, it will
be apparent to
one of ordinary skill in the art that the invention may be practiced without
these particular
details. In other instances, methods, procedures, components, and networks
that are well-
known to those of ordinary skill in the art are not described in detail to
avoid obscuring
aspects of the present invention.
[0022] Figure 1 is a block diagram illustrating an exemplary distributed
computer
system according to an embodiment of the invention. This system includes
client computer
104, aggregated reviews backend server 102, aggregated reviews frontend server
100,
multiple Web sites such as Web sites 108 and 110, and communication network(s)
106 for
interconnecting these components. Client 104 includes graphical user interface
(GUI) 112.
Aggregated reviews backend server 102 collects product reviews from multiple
Web sites
108-110, identifies particular products that are associated with particular
product reviews,
generates aggregated review information for particular products, and stores
the product
reviews and the aggregated review infonnation. Aggregated'reviews frontend
server 100
receives and responds to requests from client 104 to provide an aggregated
review for a
product and/or to search within reviews for a particular product. GUI 112
includes portions of
a plurality of reviews for a product and a search input area for entering
search terms to search
for reviews of the product that contain the search terms.
[0023] Figure 2A is a block diagram illustrating aggregated reviews backend
server
102 in accordance with one embodiment of the present invention. Server 102
typically
includes one or more processing units (CPU's) 202, one or more network or
other
communications interfaces 204, memory 206, and one or more communication buses
214 for
interconnecting these components. Server 102 optionally may include a user
interface 208
comprising a display device 210 and a keyboard 212. Memory 206 may include
high speed
random access memory and may also include non-volatile memory, such as one or
more
magnetic disk storage devices. Memory 206 may optionally include one or more
storage
devices remotely located from the CPU(s) 202. In some embodiments, the memory
206
stores the following programs, modules and data structures, or a subset or
superset thereof:
= an operating system 216 that includes procedures for handling various basic
system
services and for performing hardware dependent tasks;
4
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
= a communication module 218 that is used for connecting server 102 to other
computers (e.g., Web sites 108 and 110) via one or more communication
interfaces
204 (wired or wireless), which in turn connect the backend server 102 to one
or more
communications networks such as the Internet, other wide area networks, local
area
networks, metropolitan area networks, and so on;
= a crawling module 220 for crawling Web sites (e.g., 108 and 110), which
includes
o a URL queue 222 that stores URLs to be crawled by crawler 226;
o a link parser 224 that tells crawler 226 how to extract new links to follow
from
a fetched page;
o a crawler 226 that fetches web pages, and
o a fetched pages buffer 228 that temporarily stores web pages fetched by
crawler 226;
= a review extraction module 230 for extracting product reviews from the
information
collected by crawling module 220, which includes a review parser 232 that
extracts
content pertaining to a review (e.g., the review text, author, and date), as
well as
content that identifies the product that the review is about (e.g. the product
naine,
model number, and/or other product identifiers), from the fetched pages in
buffer 228;
= an extracted reviews buffer 234 that teinporarily stores the reviews
extracted by
review parser 232;
= a review aggregating module 236 that identifies particular products that are
associated
with particular product reviews and generates aggregated review information
for
particular products;
= an aggregated review buffer 240 that stores aggregated review information
for a
product, such as the total number of reviews for the product, an average
rating for the
product, a distribution of ratings, and/or frequently appearing phrases in the
extracted
product reviews associated with the product; one or more additional types of
aggregated review information may be stored in the aggregated review buffer
240 in
other embodiments;
= a review database 242 that stores individual reviews 244 and aggregated
reviews 246;
5
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
= a review indexer 248 that indexes the reviews (244 and 246) stored in the
review
database 242 so as produce a reviews index 250; and
= the reviews index 250, which stores the index created by review indexer 248.
[0024] The reviews index 250 maps terms (e.g., words and phrases) to reviews.
In
some embodiments, the reviews index 250 also maps other values, such as
ClusterlDs (i.e.,
product identifiers) or review author names, to the corresponding reviews.
[0025] Figure 2B is a block diagram illustrating aggregated reviews frontend
server
100 in accordance with one embodiment of the present invention. The aspects of
Figure 2B
that are analogous to those in Figure 2A are not described again here. The
memory 276 in
frontend server 100 stores the following programs, modules and data
structures, or a subset or
superset thereof:
= an operating system 216, as described above;
= a communication module 218, as described above;
= a product database 252 for storing product-related information, which
includes
o product and vendor information 254 such as product names, models,
categories, and purchasing information; and
o a product-to-review map 256 that associates products in database 252 with
their corresponding ClusterlDs (i.e., product identifiers); the ClusterIDs are
mapped by the reviews index 250 to individual reviews 244 and aggregated
reviews 246;
= reviews index 250, as described above;
= a review search module 258 that answers search requests, which includes
o a "search all reviews" application 260 for searching all reviews in reviews
database 242; and
o a "search within reviews for a product" application 262 for searching within
the set of reviews for a particular product, to identify those reviews
containing
one or more search terms; and
= a presentation module 264 for formatting aggregated reviews and search
results for
display, including
6
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
o a reviews formatter 266 for formatting individual reviews and aggregated
review information for display; and
o a reviews sorter 268 for sorting the reviews for a particular product in any
of
the following ways, or a subset or superset thereof: by date, review source
(e.g., by website), review quality, or review relevance to one or more search
terms.
[0026] The reviews database 242 and reviews index 250 are generated by the
backend
server 102, and are copied from the backend server to the frontend server 100
for responding
to user requests for product review information.
[0027] Referring to Figures 2A and 2B, each of the above identified modules
and
applications corresponds to a set of instructions for performing a function
described above.
These modules (i.e., sets of instructions) need not be implemented as separate
software
programs, procedures or modules, and thus various subsets of these modules may
be
combined or otherwise re-arranged in various embodiments. In some
einbodiinents, memory
206 or 276 may store a subset of the modules and data structures identified
above.
Furthermore, memory 206 or 276 may store additional modules and data
structures not
described above.
[0028] Although Figures 2A and 2B show servers 100 and 102 as a number of
discrete items, Figures 2A and 2B are intended more as functional descriptions
of the various
features which may be present in servers 100 and 102 rather than as a
structural schematic of
the embodiments described herein. In practice, and as recognized by those of
ordinary skill
in the art, items shown separately could be combined and some items could be
separated. For
example, some items shown separately in Figures 2A and 2B could be
implemented, on single
servers and single items could be implemented by one or more servers. The
actual number of
servers used to implement backend server 102 and frontend server 100 and how
features are
allocated among them will vary from one implementation to another, and may
depend in part
on the amount of data traffic that the system must handle during peak usage
periods as well
as during average usage periods.
Collecting Product Reviews
[0029] Figure 3 is a flowchart representing a method for collecting and
aggregating
reviews for a product according to one embodiment. The process shown in Figure
3 is
7
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
performed by aggregated reviews backend server 102 (Figures 1 and 2A). It will
be
appreciated by those of ordinary skill in the art that one or more of the acts
described may be
performed by hardware, software, or a coinbination thereof, as may be embodied
in one or
more computing systems.
[0030] Backend server 102 collects product reviews, or information containing
product reviews for a plurality of products (302).
[0031] In some embodiments, collecting information coinprises selectively
crawling
review-related Web sites or portions thereof and retrieving information
containing product
reviews. Selective crawling differs from traditional crawling in that the
crawler only follows
selected links on Web pages, rather than all links.
[0032] To initiate a crawl, crawling module 220 is given a set of starting
(sometimes
called "seed") URLs (Uniform Resource Locators, i.e., web addresses) to crawl.
[0033] Crawling module 220 puts these URLs in queue 222. Crawler 226
continuously fetches pages corresponding to the URLs in the queue 222 until
either there
more no more URLs in the queue 222, or a crawl termination condition is
reached. In some
embodiments, link parser 224 tells crawler 226 how to extract new links to
follow from a
fetched page by matching regular expression patterns. In some embodiments,
link parser 224
tells crawler 226 how to extract new links to follow from a fetched page by
using text
classification techniques. As an example of the former approach, Table 1
contains pseudo
code for a linlc parser 224 that selects which links to follow on an exemplary
website
(www.productcompany.com) for reviews of electronics products.
Table 1
Exemplary Pseudo Code for Crawling Module with Link Parser
Initialize crawler queue with seed page(s) from the www.productcompany.com
website;
For each URL in the queue, do the following {
download page from URL, and remove the URL from the queue;
extract product reviews, if any, on the page, and parse and store the
extracted product
reviews;
search page to see if any portion of the page matches any predefined pattern
in a set of
predefined patterns (contentpatternl, contentpattern2, ...);
if the page does not match any of the one or more predefmed patterns {
processing of the page is complete;}
// otherwise, continue processing the page
8
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
collect all links on the page;
process each link as follows {
if the URL of the link matches any predefined pattern in a set of predefined
patterns (linkpatternl, linkpattern2, ...) OR
if the anchor text of the link matches any predefined pattern in a set of
predefined patterns (anchorpatternl, anchorpattern2, ...), do the following: {
optional: remove unneeded fields from the URL of the link;
put the URL of the link on the crawler queue;}
} II Links that do not match any of the predefined pattern(s) are not put on
the queue
// The patterns used in the processing of the crawled pages may be specific to
the
ll website from which product review information is being extracted.
[0034] Crawler 226 outputs the crawled pages to fetched pages buffer 228. In
some
embodiments, the fetched pages buffer may contain one or more files or other
data structures.
For each page that is fetched, link parser 224 parses the page and attempts to
extract links
(URLs) to follow that, if found, are added to queue 222. As noted above, in
contrast to
traditional crawls, not all the links in a page are followed by the crawler.
Rather, only those
links that parser 224 identifies as potentially leading to review pages are
followed. For
instance, parser 224 may identify such links based on the presence of terms or
patterns in the
URLs of the linlcs, or based on the anchor text of the linlcs. Anchor text is
text contained in or
otherwise associated with a link.
[0035] The information containing product reviews can be collected from a wide
variety of sources, including web pages and forum postings. In some
embodiments, the web
pages comprise Web pages of review-related Web sites and/or product-related
Web sites.
[0036] In some embodiments, collecting information comprises receiving product
reviews in a predetermined digital format, for example as data streams from
multiple product
reviewers. The predetermined digital format of the product reviews may
comprise key/value
pairs, or data organized in predefined data structures. In some embodiments,
collecting
information comprises selecting information from a repository of information
collected by a
web crawler.
[0037] Backend server 102 automatically extracts product reviews from the
collected
information (304). In particular, review parser 232, which is analogous to
linlc parser 224,
extracts content pertaining to a review (e.g., the review text, author, and
date), as well as
content that identifies the product that the review is about (e.g. the product
name, model
number, or other features), from the fetched pages in buffer 228. The
extracted information
is stored in extracted reviews buffer 234. In some embodiments, extracted
reviews buffer
9
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
234 contains a list of key/value pairs denoting the extracted product review
information, such
as the review author, content, publisher, and so on.
[0038] The information extracted from a single web page may correspond to
multiple
reviews. In some embodiments, each extracted review found on a web page is
encoded and
stored in a respective review record in the extracted reviews buffer 234.
Table 2 lists the
fields of each review record in the extracted reviews buffer 234 in an
exemplary embodiment.
In other embodiments, each review record may have a subset of the fields
listed in Table 2,
and may also include additional fields. For example, in some embodiments, the
review
records might not include the ReviewType and CrawlDate fields. In some
embodiments, the
Rating field may be a numeric field, instead of a string field. Alternately,
in some
embodiment, each review record may include both a string Rating field and a
normalized
numeric rating field (e.g., normalized witli respect to a rating scale of 0 to
1, or 0 to 5).
[0039] In Table 2, the keyword "required" indicates a required field,
"repeated"
indicates a field that may occur more than once in a record (and in rare
cases, may have zero
occurrences), and "optional" indicates an optional field. Optional fields may
be present in
some review records, while not present in other review records, for exainple
because some
reviews may not include the optional information.
Table 2
Exemplary Fields for Individual Review Records in
Individual Reviews Buffer and Reviews Database
required int64 ReviewID // unique identifier for each review
repeated string ProductNames // one or more product names associated with the
review
required int32 ReviewType // exemplary values: editorial=l, user=2
required string Rating // product rating given by the reviewer
optional string Author // author of the review
optional string Title I/ title of the review
optional string Summary // review summary
optional string Content // review content, which is the main body of the
review
optional string Date // Review date
required int32 CrawlDate // crawl date
required string URL // URL at which review was found.
// Alternately, this may be the base URL for product
// reviews at the website at which this review was found
required string Publisher // publisher of the review (e.g., CNET,
epinions.com)
// Additional Optional Fields that may be included in some embodiments:
optional int64 ClusterID // ClusterlD of the product
optional float QualityScore // Quality Score of the review
optional int32 Year // Model year, or other year associated with the product
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
(e.g., year of first publication of a book or a movie)
[0040] The data structure fields shown in Table 2, except ReviewID,
ReviewType,
QualityScore and ClusterlD, are filled with data from the extracted review
information if
information relevant to those fields are found in the extracted review
information. The
ReviewID is a unique identifier assigned to each individual review by the
aggregated reviews
backend server. The ReviewType indicates whether the review is an editorial
review, or a
review submitted by a user. QualityS core, which provides a measure of the
quality of a
review, is determined based on the review's length and word entropy (which
measures,
among other things, the diversity of the words or terms used in the review).
As explained
below, all reviews belonging to the same product form a "cluster" and the
ClusterID identifies
the cluster to which a review belongs.
Clustering Reviews (Identifying Reviews for the Same Product)
[0041] Backend server 102 automatically identifies particular products that
are
associated with particular product reviews (306). For at least some of the
extracted product
reviews, backend server 102 identifies a particular product that is associated
with the
extracted product review.
[0042] In some embodiments, identifying a particular product that is
associated with
the extracted product review comprises associating a unique number or string
in the extracted
product review with a particular product. In some embodiments, the unique
number or string
is a product model number. In some embodiments, the unique number is an ISBN
number.
[0043] In some embodiments, identifying a particular product that is
associated with
the extracted product review comprises associating a brand name and a model
number in the
extracted product review with a particular product. In some embodiments,
identifying a
particular product that is associated with the extracted product review
comprises associating a
brand name, a model number, and a product category in the extracted product
review with a
particular product.
[0044] Different web sites often use different names for the same product,
which
makes it difficult to automatically determine whether two reviews actually
refer to the same
product. For example, different web sites refer to the same Canon scanner as:
= Canon CanoScan 7890a002 Flatbed 7890a002
= Canon CanoScan LiDE 30
11
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
= Canon CanoScan LiDE 30 Scanner
= Canon CanoScan LiDE 30 Color Scamler
= Canon Lide 30 (7890A002)
[0045] For products with global identifiers (such as ISBN numbers for books),
identifying a particular product that is associated with the extracted product
review is simple
if the product review includes the global identifier. In some embodiments, in
the absence of
such a global identifier in the review, extracted brand names and model
numbers (in the
above example, the brand name is "canon", and the model numbers are "lide30"
and
"7890a002"), as well as category information (such as the fact that the above
product is a
scanner) are used to identify reviews that refer to the same product. As used
herein, reviews
about the same product form a "cluster" and "clustering" refers to identifying
reviews for the
same product. As described above, in some embodiments, this information is
stored in
review records (Table 2) by review extraction module 230. Extracted brand
names, model
numbers and category information may all be stored in the ProductNames fields
of the review
records. Alternately, in some embodiments each review record may include a
Model field for
storing a model number, a Category field for storing product category
information, and a
Names field for storing brand name information for a respective product.
[0046] Sometimes, one or more of these three pieces of information (Model,
Category
and Brand Name) are not explicitly identified on a web page. In such cases,
this information
is deduced from the title string or otlier fields of the extracted review. For
example, product
category and model number can be inferred by the review parser 232 or the
review
aggregating module 236 as follows. Given a title string, the part of the title
string beyond
words like "for", "with", and "featuring" is ignored. Words in the title
string are traversed
from right to left and the first word (or pair of words) that is included in a
data file containing
a predefined list of product categories is considered the product category. In
some
embodiments, this data file also identifies, for each distinct product
category, a list of other
product category names for the product category that are mapped to one
canonical name. For
example, "notebooks", "notebook", "laptop" or "laptops" may all be mapped to
the canonical
name, "notebooks". Thus, a title string such as "Canon canoscan lide 30 color
scanner with
free extras" would be mapped to the category "scanner".
[0047] To extract model numbers, the title string is traversed from left to
right and the
first word that contains digits (i.e., characters from'0'-'9') is concatenated
with the preceding
word to form a model number. In some embodiments, the resulting model number
is
12
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
discarded if it matches any of the regular expression patterns corresponding
to blacklisted
model numbers or contains a brand name or parts of a brand name. An example of
a
blacklisted pattern is "pentium \d+" +" (i.e., "pentium" followed by a string
of digits). For
the title string in the previous example, "lide 30" is extracted as the model
number. Note that
more than one model number can be inferred for a product name.
[0048] In some embodiments, the brand names are identified by matching words
in
the product title with known (i.e., predefined) brand names in a list or
dictionary of such
brand names. In some embodiments, the brand name of a product may be extracted
from the
URL of the page on which the product review is found.
[0049] In some embodiments, two product reviews are assigned to the same
cluster
(product) if they have the same exact brand name, and agree on their model
numbers and
categories. Two reviews are said to agree on a model number if the set of
model numbers
extracted for them have a non-empty intersection. Two categories are said to
agree if they
are the same or at least one of them is unknown.
[0050] In some embodiments, two product reviews are assigned to the same
cluster
(product) if they have the same exact brand name and agree on their model
numbers.
[0051] In some embodiments, reviews corresponding to products for which brand
or
model information cannot be determined are put into a singleton cluster of
their own and are
not clustered with any other reviews.
[0052] In some embodiments, when a product review is assigned to a cluster,
review
aggregating module 236 fills in the ClusterlD field in the review record for
the extracted
review.
Generating Aggregate Information
[0053] For each particular product in at least a subset of the plurality of
products,
backend server 102 generates aggregated review information for the particular
product based
on a plurality of extracted product reviews that are associated with the
particular product
(308). In some embodiments, review aggregating module 236 uses the reviews for
a
particular product to generate aggregated review information for the product
that is stored in
aggregated review buffer 240.
[0054] In some embodiments, the aggregated review information for the
particular
product comprises a total number of reviews for the product, an average rating
for the
13
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
product, a distribution of the ratings for the product, and/or frequently
appearing phrases in
the extracted product reviews associated with the product. One or more
additional types of
aggregated review information may be included in the aggregated review
information in other
embodiments.
[0055] In some embodiments, the average rating is the average of weighted
ratings
from different web sites. In some embodiments, the weight of a single review
on a web site
with K reviews for the same product is 1/K. In some embodiments, the weight
given to
reviews from a particular web site or reviewer can be personalized by a user.
For example, a
user can specify via a preferences dialog box one or more web sites (e.g.,
www.consumerreports.org) and/or reviewers (e.g., Roger Ebert) preferred by the
user. The
ratings given in the reviews associated with the user's preferred sources are
then given higher
weight when computing an average rating for a product. In one embodiment, the
ratings
given in reviews associated with the user's preferred sources are given N
times the default
weight assigned to those reviews, where N is a value such as 2, 2.5, 3 or any
other
appropriate value greater than 1. An exemplary representation of the average
rating
computation is:
~ ating,
average rating a; =
Y, a;
where a; represents the weight given to each rating, rating;, of each
respective review i.
[0056] In some embodiments, frequently appearing phrases in the extracted
product
reviews associated with a particular product are computed by review
aggregation module 236
as follows. For all the phrases (e.g., unigrams, bigrams or trigrams that do
not span across
sentences) that occur in reviews associated with a product (cluster), the
number of their
weighted occurrences is counted. Occurrences in the review title or summary
are given a
higher weight. When computing or selecting frequently appearing phrases, all
stop words and
words that occur in the title of the product name are ignored. Each phrase is
given a score
equal to the length of the phrase times the square root of the sum of weighted
occurrences of
the phrase. A predetermined number of phrases with the highest scores (e.g.,
the top 6, 7, 8,
9 or 10 phrases) that do not overlap each other are chosen as the frequent
phrases and stored
in aggregated review buffer 240.
[0057] Table 3 lists the fields of each aggregated review record, produced by
aggregating information from all the extracted reviews for a particular
product, and stored in
the aggregated reviews buffer 240. In other embodiments, each aggregated
review record
14
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
may have a subset of the fields listed in Table 3, and may also include
additional fields. A
separate instance of the aggregated review record is included in the
aggregated reviews buffer
240 and in the reviews database 242 for each distinct product. In this
example, each
aggregated review record includes a ClusterlD (identifying the product
associated with this
record), a list of ReviewID's (each of which is uniquely associated with a
review of the
product), a list of the words or phrases most frequently found in the reviews
of the product,
and a computed average rating of all the reviews for the product. In some
embodiments, an
aggregated review record also includes one or more of the following additional
fields: a string
identifying the product, a number specifying the number of reviews in the
cluster; and data
concerning the distribution of ratings in the reviews for a product, such as
the number of
reviews with low, medium and high ratings, as well as the number of reviews
having no
product rating.
Table 3
Exemplary Fields for Aggregated Review Records in
Aggregated Reviews Buffer and Reviews Database
required int64 ClusterID // this is the ClusterlD for a particular product
repeated int64 ReviewID // this represents a list of ReviewID's, which are the
// IDs of the individual reviews for the product
repeated string FrequentPhrases // This is a list of phrases that frequently
occur
// in the reviews of the product
required float AvgRating // This is the average rating for the product
// Additional Optional Fields that may be included in some embodiments:
optional int32 ReviewNum // number of reviews in cluster
repeated string ProductName ll name(s) of Product to which this cluster
belongs
repeated int32 ScoreDistribution // score distribution for: unrated, negative,
neutral,
II and positive reviews
e
[0058] Backend server 102 automatically stores the extracted product reviews
244
and the aggregated review information 246 in reviews database 242 (310). In
some
embodiments, each stored product review in the reviews database 242 comprises
a record that
includes the same information as the corresponding review record (Table 2)
stored in the
extracted reviews buffer 234, while in other embodiments, the stored product
review may
comprise a subset of the information stored in the corresponding review
record.
[0059] As mentioned earlier, reviews indexer 248 indexes the individual
product
reviews 244 and the aggregated review information for a product 246 and stores
the results in
reviews index 250.
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
Displaying an aggregated review
[0060] Client 104 sends a request for an aggregated review for a product (402,
Figure 4). In some embodiments, the request is sent in response to a user-
initiated activity,
such as a user clicking on a hyperlink. For example, referring to Figure 6A, a
user looking at
web page 650 with various products for sale could activate hyperlink 652 to
see an
aggregated review for product 654.
[0061] Referring to Figure 4, frontend server 100 receives the request (404)
for an
aggregated review for a product and obtains (406) the aggregated review. In
some
embodiments, product information 254 in product database 252 is linked to
aggregated
reviews for the corresponding products by product-to-review map 256. In some
embodiments, map 256 contains mappings of product identifiers used by a first
website or
service (e.g., a product information website) to the ClusterlDs used by the
frontend server
100. The process of assigning product identifiers to products in the first
website or service
may be analogous to the process described above for identifying particular
products that are
associated with particular product reviews. In some embodiments, map 256 is
determined by
matching a global identifier for a particular product (e.g., an ISBN number)
in product
database 252 with a corresponding review cluster. Product-to-review map 256 is
used to
identify the aggregated review (in the reviews database 242) corresponding to
a product.
Presentation module 264 formats the aggregated review. In some embodiments,
the reviews
index 250 is used to identify the individual reviews corresponding to the
aggregated review
for a product. More specifically, the reviews index 250 maps the ClusterlD of
the aggregated
review to the individual reviews for the product associated with the
aggregated review.
[0062] Frontend server 100 sends (408) the aggregated review to client
computer 104
via communications network 106. Client 104 receives (410) and displays (412)
the
aggregated review 600 (Figure 6B).
[0063] As shown in Figure 6B, aggregated review 600 includes portions of
reviews
626 for the product and a search input area 618 for entering search terms to
search for
reviews of the product that contain the search terms. Search input area 618 is
used to enter
search terms for a search that only returns results that are within the
reviews for the product.
In some embodiments, aggregated review 600 includes a histogram 608 of review
ratings for
the product, an aggregate rating 604 for the product, and/or the number 606 of
reviews for the
product. In some embodiments, aggregated review 600 is displayed in a single
Web page.
16
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
Searching
[0064] In some embodiments, review search module 258 supports two different
types
of queries: (1) search within the reviews 262 belonging to a single product
(e.g. a search for
"customer service" in the reviews for a given product) and (2) search all
reviews 260 to find a
particular product (e.g. a search for "good digital camera"). Aggregated
review 600, Figure
6B, shows a search input area 618 for the first type of search.
[0065] For the first type of search, a variety of user initiated activities
can start the
search process, including the user entering search terms in search input area
618, selecting a
hyperlink for a frequently appearing phrase 620, selecting a hyperlink to a
review source 622,
or selecting a hyperlinked portion (e.g., 610, 612, 614, or 616) of histogram
608.
[0066] In response to this user-initiated activity, client 104 sends (502,
Figure 5) a
request to search within a plurality of reviews for a particular product.
[0067] Referring to Figure 5, frontend server 100 receives (504) the search
request
from client 104 and obtains (506) information satisfying the request. In some
embodiments,
the information is obtained by review search module 258 sending two queries to
reviews
database 242. In the first query, all individual reviews (sometimes called
"documents") that
contain the search terms are found. When searclling for reviews within a
single product, the
first query is restricted to the main body (e.g., the "review content" fields)
of the reviews
index 250, and only to individual reviews that belong to the product
(cluster). For example,
the query "customer service" for reviews in cluster 1234 would be translated
to the following:
query restricts: ClusterlD=1234, body:(customer service).
[0068] A second query is then issued to obtain portions of all reviews that
are to be
displayed. The ordering preferences submitted by the user (e.g., by clicking
on 636, 638,
640, and/or 642) as well as the start position are used to determine which
reviews to display.
When the number of reviews for a product exceeds the maximum number of review
portions
to be displayed on a single web page, the start position indicates the ordinal
number within an
ordered set of reviews at which to begin formatting reviews for display.
[0069] Frontend server 100 sends (508, Figure 5) at least a portion of the
obtained
information to client 104.
17
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
[0070] Client 104 receives (510) information satisfying the request and
displays (512)
the received information, e.g. in web page 660, Figure 6C. In some
embodiments, Client 104
displays at least portions 662 of reviews of the product matching the search
terms 664.
[0071] The information sent by frontend server 100 and received by client 104
includes at least portions 662 of reviews for the particular product that
satisfy the search
request. In some einbodiments, this information includes the identities 666 of
the review
sources contributing to the sent information.
[0072] For the second type of search, the information is also obtained by
review
search module 258 sending two queries to reviews database 242. In the first
query, all
individual reviews that contain the search terms are found, without
restricting the search to a
particular product (cluster). In some embodiments, the results of the first
query are grouped
by cluster (product). A second query is then issued to obtain portions of at
most three (or
some other small number) reviews in each cluster, which are restricted to be
from different
publishers. In some embodiments, information about each of the clusters, such
as the total
number of reviews and the average rating, is also requested, obtained, and
sent for display.
Graphical User Interface
[0073] The graphical user interface 600 (Figure 6B) on client computer 104
comprises a plurality of portions 626 of reviews for a product and a search
input area 618 for
entering search terms to search for reviews of the product that contain the
search terins. In
some embodiments, the plurality of portions of reviews can be grouped by
review source 636,
sorted by date 642, and sorted by relevance to the search input 640, or sorted
by quality of
review when the user has not provided a search input. In some embodiments,
more or fewer
sorting and grouping options may be provided for sorting and/or grouping the
reviews.
[0074] In some embodiments, graphical user interface 600 also includes, either
alone
or in various combinations: an aggregate rating 604 for the product; a list of
suggested search
terms 620; a list of sources 622 for the reviews that link back to the source
Web sites or to the
corresponding reviews; at least one user selectable link for sorting the
portions of reviews by
date 642 or relevance to the search input 640, and/or for grouping the
portions of reviews by
source 636; and/or a histogram 608 of reviews for the product. In some
embodiments,
clicking on a region (e.g., 610, 612, 614, or 616) in the histogram
automatically generates
portions of at least some of the reviews corresponding to the histogram
region. For example,
18
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
clicking on region 616 may generate a new display showing portions of reviews
having high
ratings for the product.
[0075] An aggregated rating 604 is given to the product by calculating an
average
over all reviews for the product. The ratings in the individual reviews 244
and in the
aggregated reviews 246 are normalized, e.g., to a 0-1 or 0-5 scale. In some
embodiments,
after a search within reviews for a product 262, the aggregated rating 604 and
number 606 of
reviews are changed to reflect the average rating and number of reviews
matching the search.
This change in the aggregated rating can help users answer questions like,
"How do reviewers
that discuss 'customer service' rate this product?" In some other embodiments,
the aggregated
rating 604 and number 606 of reviews are not changed to reflect the average
rating and
number of reviews matching the search.
[0076] Commonly used phrases (e.g., unigrams, bigrams or trigrams) appearing
in the
reviews for the product are displayed as a list of suggested search terms 620.
Phrases that
contain the product name and stop-words (such as "the") are removed to improve
the quality
of the suggested search terms. Clicking on one of the phrases results in
displaying just the
reviews that contain the user selected phrase. This is equivalent to entering
the phrase into
the search box 618 with quotes around the phrase and then initiating a search.
Showing these
phrases gives the user a good idea which aspects of the product are most
frequently discussed
in the product reviews, and thus which aspects of the product might be
important to lcnow
about.
[0077] The list of review sources 622 includes links to the sites from which
reviews
for the product have been obtained, or a subset thereof. In some embodiments,
after a search
within reviews for a product, the list of review sources 622 just includes
links to the sites
from which reviews matching the search have been obtained, or a subset
thereof. In some
embodiments, the linlcs lead to the main pages for the given product on each
source web site.
In some embodiments, the links lead to the product reviews matching the
search.
[0078] A histogram 608 with a rating distribution of the reviews includes the
number
of reviews that are unrated 610 (in some embodiments, displayed as a gray
area), negative
612 (in some embodiments, displayed as a red area), neutral 614 (in some
embodiments,
displayed as a yellow area) and positive 616 (in some embodiments, displayed
as a green
area). In some embodiments, the rating distribution is displayed as a flat
histogram, where
the length of each colored bar is proportional to the number of reviews in the
category.
19
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
Clicking on a histogram results in a search 262 that just shows the
corresponding reviews,
e.g. clicking on the red bar 612 results in showing just the negative reviews.
[0079] Search input area 618 allows a user to enter arbitrary search terms for
the
search within reviews for a given product. As shown in Figure 6C, in some
einbodiments the
search terms are highlighted in portions of reviews 626 that match the search
query.
[00801 User selectable linlcs permit the displayed list of reviews to be
sorted in
various ways, such as by:
Relevance 640: In some embodiments, editorial reviews are listed before
user reviews. In some embodiments, the reviews are sorted by their length
and information content, e.g., sorted by Review-per-word-entropy *
sqrt(Review-Length). Other scoring functions, such as scoring reviews by
an "information retrieval" score, can also be used.
Date 642: The reviews are sorted by date, with the newest reviews coming
first. Reviews for which a date is unknown may be listed at the end.
Review Source 636: Reviews are grouped by review site, as represented
by screen shot 670 in Figure 6D, with at most N reviews (e.g., 3, 4 or 5, or
some other small number) shown for each review site. At the top of each
group, a link 672 to the corresponding review site is provided, which may
be the same as the link provided to the site in the list of review sources
622.
Ungrouped 638 (e.g., not grouped by review source). This view is similar
to traditional search result displays. Reviews are listed (i.e., portions of
the
reviews are displayed) one-by-one in a selected order (e.g., by date or
relevance) and reviews from different sources are mixed (i.e., without
grouping the portions of reviews by source).
Location (not shown in aggregated review 600, Figure 6B). This view
permits reviews to be sorted by location (e.g., for reviews of restaurants,
local businesses, movies or other products near a particular location, such
as by address, city, region, or mailing or zip code).
CA 02591441 2007-06-13
WO 2006/065546 PCT/US2005/043459
[0081] In some embodiments, for reviews listed in an aggregated review (e.g.,
graphical user interface 600 or 670) or reviews listed as results for a search
within reviews for
a product (e.g., graphical user interface 660, Figure 6C) the following
information is
displayed: review title 624, a portion 626 of the review (e.g., a snippet),
the source of the
review 628, the date of the review 634, the type of review 630 (e.g., whether
it is an editorial
or user contributed review), and the name of the reviewer 632. The portion or
snippet 626 of
each review may be generated using the same or similar teclmiques to those
used to generate
snippets of search results in an Internet search engine such as the Google
search engine. For
example, when a graphical user interface is presenting reviews for a product,
the portion of
each review shown may be a portion matching one or more search terms that were
used to
select the reviews to be displayed. Alternately, the portion of each review
shown may be a
beginning portion of the review. In yet another embodiment, the portion of
each review
shown may be a portion determined by an automatic process to be a summary or
most
important portion of the review.
[0082] The foregoing description, for purpose of explanation, has been
described with
reference to specific embodiments. However, the illustrative discussions above
are not
intended to be exhaustive or to limit the invention to the precise forms
disclosed. Many
modifications and variations are possible in view of the above teachings. The
embodiments
were chosen and described in order to best explain the principles of the
invention and its
practical applications, to thereby enable others skilled in the art to best
utilize the invention
and various embodiments with various modifications as are suited to the
particular use
contemplated.
21
