U.S. patent application number 14/952201 was filed with the patent office on 2016-04-28 for method and system for determining on-line influence in social media.
This patent application is currently assigned to salesforce.com, inc.. The applicant listed for this patent is salesforce.com, inc.. Invention is credited to Marcel Albert Lebrun, Christopher Daniel Newton, Christopher Bennett Ramsey.
Application Number | 20160117696 14/952201 |
Document ID | / |
Family ID | 41267616 |
Filed Date | 2016-04-28 |
United States Patent
Application |
20160117696 |
Kind Code |
A1 |
Newton; Christopher Daniel ;
et al. |
April 28, 2016 |
METHOD AND SYSTEM FOR DETERMINING ON-LINE INFLUENCE IN SOCIAL
MEDIA
Abstract
Methods and system are provided for determining a topical
influence value of the individual based on aggregated viral
properties of tagged content citing the individual. A processor of
a computer is used to match content within a web-site with a
selected topic and to tag matching content to generate tagged
content. Viral properties for the tagged content are extracted, and
viral properties of the tagged content citing an individual in the
tagged content are aggregated to form aggregated viral properties
of the tagged content citing the individual. Based on the
aggregated viral properties of the tagged content citing the
individual, a topical influence value of the individual can be
computed.
Inventors: |
Newton; Christopher Daniel;
(Douglas, CA) ; Lebrun; Marcel Albert;
(Fredericton, CA) ; Ramsey; Christopher Bennett;
(Fredericton, CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
salesforce.com, inc. |
San Francisco |
CA |
US |
|
|
Assignee: |
salesforce.com, inc.
San Francisco
CA
|
Family ID: |
41267616 |
Appl. No.: |
14/952201 |
Filed: |
November 25, 2015 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
12437418 |
May 7, 2009 |
9245252 |
|
|
14952201 |
|
|
|
|
12174345 |
Jul 16, 2008 |
|
|
|
12437418 |
|
|
|
|
61051202 |
May 7, 2008 |
|
|
|
Current U.S.
Class: |
705/7.29 |
Current CPC
Class: |
G06Q 50/01 20130101;
G06Q 10/10 20130101; G06Q 30/0201 20130101 |
International
Class: |
G06Q 30/02 20060101
G06Q030/02; G06Q 50/00 20060101 G06Q050/00 |
Claims
1. A computer-implemented method performed by a computer in
communication with a server via network to access a web-site hosted
by the server, the computer-implemented method comprising steps of:
using a processor of the computer to match content within the
web-site with a selected topic and to tag matching content to
generate tagged content; extracting, with the processor, viral
properties for the tagged content; aggregating, with the processor,
the viral properties of the tagged content citing an individual in
the tagged content to form aggregated viral properties of the
tagged content citing the individual; and computing, with the
processor, a topical influence value of the individual based on the
aggregated viral properties of the tagged content citing the
individual.
2. The computer-implemented method of claim 1, further comprising
the steps of: aggregating, with the processor, the viral properties
for the tagged content from the web-site to form aggregated viral
properties of the tagged content from the web-site; and computing,
with the processor, a topical influence value of the web-site based
on a combination of the aggregated viral properties of the tagged
content from the web-site, and the topical influence value of the
individual cited in the tagged content.
3. The computer-implemented method of claim 2, further comprising:
repeating the steps of matching and tagging, extracting,
identifying, aggregating and computing, the repeating of the
computing step including: updating the topical influence value of
the individual to generate an updated topical influence value of
the individual; and updating the topical influence value of the
web-site to generate an updated topical influence value of the
web-site by integrating the updated topical influence value of the
individual in the computation of the updated topical influence
value of the web-site.
4. The computer-implemented method of claim 1, wherein the step of
computing the topical influence value of the individual, further
comprises: identifying an outlet owned by the individual;
calculating a topical influence value of the outlet owned by the
individual based on a combination of viral properties extracted
from the outlet owned by the individual; and updating the topical
influence value of the individual based on a combination of the
aggregated viral properties of the tagged content citing the
individual and the topical influence value of the outlet owned by
the individual.
5. The computer-implemented method as described in claim 1, further
comprising: iterating the steps of matching and tagging,
extracting, identifying, aggregating and computing for a plurality
of individuals; and identifying top influential individuals from
the plurality of individuals based on the topical influence values
for each of the plurality of individuals.
6. The computer-implemented method as described in claim 1, wherein
the step of extracting viral properties for the tagged content
comprises: collecting values of the viral properties at
predetermined time intervals; and storing the collected values in
respective time series.
7. The computer-implemented method of claim 1, wherein the viral
properties are selected from the group consisting of: user
engagement value; cited individual count, inbound links; sub
scribers; average Social bookmarks; average Social news votes;
buries; total count of posts; and total count of appearance of
individuals names across all posts.
8. A method performed by a computer for in communication with a
server via network to access a web-site hosted by the server, the
method comprising the steps of: using a processor of the computer
to match content within the web-site with a selected topic and to
tag matching content to generate tagged content; extracting, with
the processor, viral properties for the tagged content;
aggregating, with the processor, the viral properties across all
tagged content contained within the web-site including the viral
properties of the tagged content citing the individual in the
tagged content to form aggregated viral properties of the tagged
content citing the individual; computing, with the processor, the
topical influence value of the individual based on a combination of
the aggregated viral properties of the tagged content citing the
individual; and repeating the steps of matching and tagging,
extracting, identifying, aggregating and computing, the repeating
of the computing step including: updating the topical influence
value of the individual to generate an updated topical influence
value of the individual.
9. The method of claim 8, further comprising: iterating the steps
of matching and tagging, extracting, identifying, aggregating and
computing for a plurality of individuals; and identifying top
influential individuals from the plurality of individuals based on
the topical influence values for each of the plurality of
individuals.
10. The method of claim 8, wherein the step of computing the
topical influence value of the individual comprises: identifying an
outlet owned by the individual; determining an influence of the
outlet characterized by a topical influence value of the outlet;
and calculating the topical influence value of the individual by
including the topical influence value of the outlet in the
combination of the aggregated viral properties of the tagged
content citing the individual.
11. The method of claim 8, further comprising: aggregating, with
the processor, the viral properties for the tagged content from the
web-site to form aggregated viral properties of the tagged content
from the web-site; computing, with the processor, the topical
influence value of the web-site based on a combination of the
aggregated viral properties of the tagged content from the
web-site, and the topical influence value of the individual cited
in the tagged content, the topical influence value of the web-site
characterizing the topical on-line influence of the web-site; and
repeating the steps of matching and tagging, extracting,
identifying, aggregating and computing, the repeating of the
computing step including: updating the topical influence value of
the individual to generate an updated topical influence value of
the individual; and updating the topical influence value of the
web-site to generate an updated topical influence value of the
web-site by integrating the updated topical influence value of the
individual in the computation of the updated topical influence
value of the web-site.
12. The method of claim 11, further comprising: iterating the steps
of matching and tagging, extracting, identifying, aggregating and
computing for a plurality of web-sites; and identifying top
influential sites based on the topical influence values for each of
the plurality of web-sites.
13. A non-transitory computer readable medium, comprising computer
code instructions stored thereon, which, when executed by a
computer coupled to a network and in communication with a server
via the network to access a web-site hosted by the server, cause
the computer to perform the steps of: matching content within the
web-site with a selected topic and tagging matching content to
generate tagged content from the web-site; extracting viral
properties for the tagged content; aggregating the viral properties
of the tagged content citing an individual in the tagged content to
form aggregated viral properties of the tagged content from the
web-site citing the individual; and computing a topical influence
value of the individual based on a combination of the aggregated
viral properties of the tagged content citing the individual.
14. The non-transitory computer readable medium of claim 13,
comprising the computer code instructions stored thereon, which,
when executed by the computer, further cause the computer to
perform the steps of: aggregating the viral properties for the
tagged content from the web-site to form aggregated viral
properties of the tagged content from the web-site; and computing a
topical influence value of the web-site based on a combination of
the aggregated viral properties of the tagged content from the
web-site, and the topical influence value of the individual cited
in the tagged content contained within the web-site.
15. The non-transitory computer readable medium of claim 14,
comprising the computer code instructions stored thereon, which,
when executed by the computer, further cause the computer to
perform the steps of: repeating the steps of matching and tagging,
extracting, identifying, aggregating and computing, the repeating
of the computing step including: updating the topical influence
value of the individual to generate an updated topical influence
value of the individual; and updating the topical influence value
of the web-site to generate an updated topical influence value of
the web-site by integrating the updated topical influence value of
the individual in the computation of the updated topical influence
value of the web-site.
16. The non-transitory computer readable medium of claim 12,
comprising the computer code instructions stored thereon, which,
when executed by the computer, further cause the computer to
perform the steps of: identifying an outlet owned by the
individual; calculating a topical influence value of the outlet
owned by the individual based on a combination of viral properties
extracted from the outlet owned by the individual; and updating the
topical influence value of the individual based on a combination of
the aggregated viral properties of the tagged content citing the
individual and the topical influence value of the outlet owned by
the individual.
17. The non-transitory computer readable medium of claim 12,
comprising the computer code instructions stored thereon, which,
when executed by the computer, further cause the computer to
perform the steps of: iterating the steps of matching and tagging,
extracting, identifying, aggregating and computing for a plurality
of individuals; and identifying top influential individuals from
the plurality of individuals based on the topical influence values
for each of the plurality of individuals.
18. A system, comprising: a network; a server coupled to the
network, wherein the server hosts the web-site; a computer coupled
to the network and in communication with the server to access the
web-site, the computer comprising: a processor; and a computer
readable storage medium having computer readable instructions
stored thereon for execution by the processor, forming the
following modules: a content-to-topic matching module for matching
content within the web-site with a selected topic and tagging
matching content to generate tagged content from the web-site; a
viral properties extraction module for: extracting viral properties
of the tagged content; and aggregating the viral properties of the
tagged content citing the individual to form aggregated viral
properties of the tagged content citing the individual; an
individual influence modeling module for computing, based on a
combination of the aggregated viral properties extracted from the
tagged content citing the individual, a topical influence value of
the individual cited in the tagged content; and a database
comprising: a top influential individuals data store for storing
the topical influence value.
19. The system of claim 18, wherein the computer readable storage
medium having computer readable instructions stored thereon for
execution by the processor further forms: a site influence modeling
module for computing the topical influence value of the web-site
based on a combination of the aggregated viral properties of the
tagged content within the web-site, and the topical influence value
of the individual cited in the tagged content, and wherein the
database further comprises: a top influential sites data store for
storing the topical influence value of the web-site.
20. The system of claim 18, wherein the individual influence
modeling module further comprises: an individual influence updating
module for updating the topical influence of the individual by
calculating the topical influence value of the individual based on
viral properties extracted from tagged content identifying the
individual and topical influence values of outlets owned by the
individual.
Description
RELATED APPLICATIONS
[0001] The present application is a Continuation of the U.S.
application Ser. No. 12/437,418, filed May 7, 2009, which is a
Continuation-in-Part (CIP) of the U.S. application Ser. No.
12/174,345 filed Jul. 16, 2008 entitled "Method and System for
Determining Topical On-Line Influence of an Entity", and claims
benefit from the US provisional application to Christopher NEWTON
Ser. No. 61/051,202 filed on May 7, 2008 entitled "Method and
System for Determining On-Line Influence in Social Media", and the
U.S. application Ser. No. 12/174,345 filed Jul. 18, 2008 entitled
"Method and System For Determining Topical On-Line Influence of an
Entity", both of which are incorporated herein by reference.
FIELD OF THE INVENTION
[0002] The present patent application relates to a computer
implemented method and system for determining on-line influence in
social media, and in particular, to a computer implemented method
and system for determining on-line influence of a commentor, an
individual, or a web-site.
BACKGROUND OF THE INVENTION
[0003] Determining on-line influence of a commentor, an individual,
an entity, or a web-site, also to be referred herein as site, in
social media becomes an increasingly important subject nowadays. A
major problem facing marketers and public relation (PR)
professionals revolves around the prolific use of social media
sites and the awesome scale they have achieved. Literally, hundreds
of thousands of videos, blog posts, podcasts, events, and social
network interactions, such as wall posts, group postings, and
others, occur daily. Due to the sheer volume of content, constantly
changing landscape of popular sites, and hundreds of millions of
users involved, it is impossible to determine who should be
listened to and those who must be engaged.
[0004] Existing systems for determining influence in social media
are site based, i.e. their models of influence are calculated on a
per-site basis. If there is a one-to-one relationship between a
site and a person (an author), then the influence is extrapolated
to indicate the influence of the person for the medium, in which
the site exists. For instance, if siteA is a blog with only one
author, and all blogs are counted similarly, then the influence for
siteA, as calculated by the prior art methods, would also indicate
the influence for the author of the blog.
[0005] To the best of the knowledge of inventors, prior art methods
predominantly calculate influence in social media by recursively
analyzing inbound web page link counts. For example, siteA would
have a higher influence score than siteB if the following
approximate rules apply:
RULE 1: If the number of sites having links pointing at siteA is
higher than the number of sites having links pointing at siteB; and
RULE 2: If the count of sites pointing at the sites that point at
siteA is higher than the count of sites that are pointing at the
sites that point at siteB.
[0006] Rule 2 is applied recursively.
Various issues exist with this prior art method, namely: [0007] The
method assumes that the total influence of the sites can be
measured by a single property and that no other factors affect
influence to a scale large enough to invalidate using only inbound
link count as the measured property; [0008] The method assumes that
the link graph representing all the links between the sites is
complete enough to form a basis for determining an influence;
[0009] The method assumes that a link implies that the linker has
been influenced by the site he is linking to, which is not
necessarily the case; [0010] The method does not account for
connections someone may have with a site, if there is no link to
track that connection, i.e. if a visitor does not own a blog, and
therefore does not link out to anyone, but he is still a frequent
visitor to the blog, e.g., http://www.autoblog.com, then the
influence that Autoblog has over the visitor is not calculated; and
[0011] The method does not map properly to other types of content
and methods of social media expression, e.g., link-analysis methods
deployed to the blogosphere are not relevant in the micromedia
sphere of Twitter, i.e. link analysis techniques do not translate
to all forms of social media and therefore they leave out entire
pools of influencers that use other media channels as their
voice.
[0012] Accordingly, there is a need in the industry for developing
alternative and improved methods and system for determining on-line
influence of individuals and commentors cited or published in
social media content as well as for determining the influence of
web-sites hosting the content.
SUMMARY OF THE INVENTION
[0013] There is an object of the invention to provide an improved
method and system for determining topical on-line influence of a
commentor, an individual, or a site, which would avoid or mitigate
the above mentioned drawbacks of the prior art.
[0014] According to the embodiments of the invention, a topical
on-line influence is introduced, which is a measure of how many
people are engaged in a message of an entity content site around a
given topic. Additionally the influence can also be a measure of
the popularity level of a commentor, or an individual, or how
influential they are around a given topic.
[0015] A user is allowed to manipulate equations by adding
additional viral properties to the equations or removing certain
viral properties from the equations, and by adjusting weights in
the equations.
[0016] According to one aspect of the invention, there is provided
a method for determining a topical on-line influence of a
commentor, the method comprising: [0017] (a) matching and tagging
content with a selected topic to generate tagged content; [0018]
(b) extracting viral properties for the tagged content; [0019] (c)
identifying a commentor from said tagged content; and [0020] (d)
determining the topical on-line influence of said commentor;
wherein said topical on-line influence of said commentor is
characterized by a topical influence value of the commentor; and
said topical influence value of the commentor is calculated from a
commentor influence model based on the extracted viral
properties.
[0021] Additionally, the commentor influence model is a linear
combination of the extracted viral properties weighted with
respective weights applied to each of the extracted viral
properties.
[0022] Furthermore, step (d) of the method further comprises:
[0023] determining topical influences of on-line outlets owned by
said commentor said topical influences characterized by respective
topical influence values of the on-line outlets; and [0024]
creating an updated commentor influence model by forming a linear
combination of the topical influence values of said on-line outlets
and the topical influence value of the commentor.
[0025] Beneficially, each topical influence of the topical
influences of the on-line outlets owned by said commentor is
determined from a linear combination of viral properties extracted
from a corresponding on-line outlet.
[0026] In one modification, the method further comprises: [0027]
(e) aggregating the viral properties of tagged content contained
within a web-site and having a comment from said commentor to form
aggregated viral properties of the tagged content within the
web-site; and [0028] (f) calculating a topical influence value of
the web-site based on a linear combination of the aggregated viral
properties of the tagged content within the web-site and the
topical influence value of the commentor.
[0029] In a further modification, the method comprises: [0030] (g)
identifying an individual cited in the tagged content contained
within the web-site; [0031] (h) aggregating viral properties of the
tagged content citing said individual to form aggregated viral
properties of the tagged content citing said individual; and [0032]
(i) calculating an influence value of said individual based on a
linear combination of the aggregated viral properties of the tagged
content citing said individual.
[0033] In another modification, the method further comprises:
[0034] (j) updating the topical influence value of the web-site by
calculating the topical influence value of the web-site based on a
linear combination of the aggregated viral properties of the tagged
content within the web-site, the topical influence value of the
commentor and the influence value of said individual.
[0035] Step (i) further comprises: [0036] (i) identifying an outlet
owned by said individual; [0037] (ii) calculating an influence
value of said outlet owned by said individual based on a linear
combination of viral properties extracted from said outlet owned by
said individual; and [0038] (iii) updating the influence value of
said individual based on a linear combination of said aggregated
viral properties of the tagged content citing said individual and
the influence value of said outlet owned by said individual.
[0039] Step (b) of the method comprises: [0040] collecting values
of the viral properties at predetermined time intervals; and [0041]
storing the collected values in respective time series.
[0042] Furthermore, the viral properties are selected from the
group consisting of: [0043] user engagement value; average comment
count; average unique commentor count; [0044] cited individual
count; inbound links; subscribers; average social bookmarks; [0045]
average social news votes; buries; total count of posts; and total
count of appearance of Individuals names across all posts.
[0046] According to another aspect of the invention, a method for
determining a topical on-line influence of a web-site is disclosed.
The method comprises: [0047] (a) matching and tagging content with
a selected topic to form tagged content; [0048] (b) extracting
viral properties for the tagged content; [0049] (c) aggregating the
viral properties across all tagged content contained within the
web-site; and [0050] (d) calculating an influence value of the
web-site based on a linear combination of the aggregated viral
properties, said influence value characterizing the topical on-line
influence of the web-site.
[0051] The method further comprises: [0052] (e) identifying a
commentor having a comment in said tagged content contained within
the web-site; [0053] (f) determining a topical influence of said
commentor wherein said topical influence is characterized by an
influence value of the commentor and said influence value of the
commentor is calculated based on a linear combination of the
aggregated viral properties of the tagged content contained within
the web-site and having the comment from said commentor; and [0054]
(g) updating the topical on-line influence of the web-site by
integrating the influence value of the commentor in the calculation
of the influence value of the web-site.
[0055] Additionally, the method further comprises: [0056] (h)
identifying an individual cited in said tagged content contained
within the web-site; [0057] (i) determining a topical influence of
said individual characterized by an influence value of the
individual and said influence value of the individual is calculated
based on a linear combination of the aggregated viral properties of
the tagged content contained within the web-site and citing said
individual; and [0058] (j) updating the topical on-line influence
of the web-site by integrating the influence value of the commentor
and the influence value of the individual in the calculation of the
influence value of the web-site.
[0059] The step of calculating the influence value of said
commentor comprises: [0060] (i) identifying an outlet owned by said
commentor; [0061] (ii) determining an influence of said outlet
characterized by an influence value of said outlet; and [0062]
(iii) calculating the influence value of said commentor by
including the influence value of said outlet in the linear
combination of the aggregated viral properties of the tagged
content contained within the web-site and having a comment from
said commentor.
[0063] The step of calculating the influence value of said
individual comprises: [0064] (i) identifying an outlet owned by
said individual; [0065] (ii) determining an influence of said
outlet characterized by an influence value of said outlet; and
[0066] (iii) calculating the influence value of said individual by
including the influence value of said outlet in the linear
combination of the aggregated viral properties of the tagged
content contained within the web-site and citing said
individual.
[0067] In one modification, the method further comprises iterating
steps (a) to (d) for a plurality of web-sites and identifying top
influential sites based on the influence value of each of the
plurality of web-sites.
[0068] In another modification, the method further comprising
iterating steps (e) to (g) for a plurality of commentors and
identifying top influential commentors based on the influence value
of each of the plurality of commentors.
[0069] In yet another modification, the method further comprises
iterating steps (h) to (j) for a plurality of individuals and
identifying top influential individuals based on the influence
value of each of the plurality of individuals.
[0070] In yet another aspect of the invention, a computer readable
medium is disclosed. The computer readable medium comprises a
computer code instructions stored thereon, which, when executed by
a computer, perform the steps of the methods of the present
invention.
[0071] In a further aspect, a system for determining a topical
on-line influence of a web-site is disclosed. The system comprises:
[0072] a processor; and a computer readable storage medium having
computer readable instructions stored thereon for execution by the
processor, forming the following modules: [0073] (a) a matching
module for matching and tagging content to a selected topic to
generate tagged content; [0074] (b) a viral properties extraction
module for extracting viral properties of said tagged content;
[0075] (c) a commentor influence modeling module for calculating an
influence value of a commentor based on viral properties extracted
from tagged content having a comment from the commentor, said
influence value of the commentor characterizing a topical influence
of the commentor; and [0076] (d) a site influence modeling module
for calculating an influence value of a web-site based on viral
properties extracted from tagged content contained in said
web-site, said influence value of the web-site characterizing a
topical influence of the web-site.
[0077] The site influence modeling module further comprises a site
influence updating module for updating the influence value of the
web-site based on said viral properties extracted from the tagged
content contained in said web-site, an influence value of a
commentor having a comment in said tagged content and an influence
value of an individual cited in said tagged content.
[0078] Additionally, the system further comprises: [0079] an
individual influence modeling module for calculating the influence
value of the individual cited in said tagged content based on viral
properties extracted from tagged content identifying said
individual, said influence value of the individual characterizing a
topical influence of the individual.
[0080] The commentor influence modeling module further comprises a
commentor influence updating module for updating the influence of
the commentor by calculating the influence value of the commentor
based on viral properties extracted from the tagged content having
a comment from the commentor and influence values of outlets owned
by said commentor.
[0081] The individual influence modeling module further comprises
an individual influence updating module for updating the topical
influence of the individual by calculating the influence value of
the individual based on viral properties extracted from tagged
content identifying said individual and influence values of outlets
owned by said individual.
BRIEF DESCRIPTION OF THE DRAWINGS
[0082] Embodiments of the invention will now be described, by way
of example, with reference to the accompanying drawings in
which:
[0083] FIG. 1 illustrates a system architecture, in which the
embodiments of the present invention have been implemented;
[0084] FIG. 2 illustrates a Content-to-Topic Matching block 150 of
the system of FIG. 1;
[0085] FIG. 3 shows a flowchart 300 illustrating the operation of
the Content-to-Topic Matching block of FIG. 2;
[0086] FIG. 4 shows a block diagram for a system for a recursive
determining of influence of a commentor, individual, or a site
according to the embodiment of the invention; and
[0087] FIG. 5 shows a flowchart illustrating the operation of the
system of FIG. 4.
DETAILED DESCRIPTION OF THE EMBODIMENTS OF THE INVENTION
[0088] Embodiments of the invention describe models for determining
an influence of a commentor, an individual, a web-site, or site,
wherein the model for determining the influence of the site uses,
recursively, an influence value for a commentor or an influence
value for an individual, or both, when the commentor or individual
is present in the site. The influence values for the commentor and
individual characterize, respectively, the influence of the
commentor and individual. The influences are topically relevant,
and can be aggregated across various forms of content. The
influence of the commentor is calculated from an influence model
for the commentor which is based on aggregated viral properties
across content having comments from the commentor. The viral
properties will be also referred to as herein as viral dynamics.
Additionally, viral properties of social media outlets owned by the
commentor, when available, are integrated into the influence model
for the commentor. The influence of the individual is calculated
from an influence model for the individual, which is based on
aggregated viral properties of the content citing the individual.
The influence model for the site hosting the topically-relevant
content is based on the extracted viral properties of the content.
Further fine-tuning of the influence model for the site is provided
by selectively integrating the influence value for the commentors
and influence value for the individuals.
[0089] FIG. 1 illustrates architecture of a system 100, in which
the embodiments of the present invention have been implemented. The
system 100 includes a Content-to-Topic Matching Module 150, an
Influence Calculation & Viral Properties Extraction Module 160
and a Recursive Influence Modeling Module 110. The system 100
includes also Surfaced Top Influential Sites data store 120,
Surfaced Top Influential Commentors data store 130 and Surfaced Top
Influential Individuals data store 140.
[0090] As shown in FIG. 1, the Recursive Influence Modeling Module
110 is connected to the Content-to-Topic Matching Module 150. The
Recursive Influence Modeling Module 110 is also connected to the
Influence Calculation & Viral Properties Extraction Module 160
and to the data stores 120, 130 and 140. The Recursive Influence
Modeling Module 110 processes information received from the
Content-to-Topic Matching Module 150 and the Influence Calculation
& Viral Properties Extraction Module 160, and generates, as
output, data related to top influencers.
[0091] Surfaced Top Influential Sites data store 120, Surfaced Top
Influential Commentors data store 130 and Surfaced Top Influential
Individuals data store 140 contain, respectively, top influential
sites, top influential commentors and top influential individuals.
The data in the data stores 120, 130 and 140 is generated from the
Recursive Influence Modeling Module 110. The data stores 120, 130
and 140 are implemented as databases stored in a computer readable
storage medium, e.g., computer memory, DVD, CD-ROM, accessible
through a database query language. Other alternative
implementations can also be adopted to store and retrieve the
stored data.
[0092] The system illustrated in FIG. 1 comprises a computer,
having a processor and a computer readable storage medium having
computer readable instructions stored thereon for execution by the
processor, and forming modules of the system of FIG. 1 described
above. The computer readable storage medium can be implemented as a
computer memory, CD-ROM, DVD, floppy, hard drive or the like. The
modules of the system of FIG. 1, when executed by the computer,
perform the steps of the method of the embodiment of the present
invention.
[0093] As an example, each of the individual modules of the system
100 can be implemented as an individual software module or agent
running on a common hardware platform. Other alternative
implementations are possible and are well known to the persons
skilled in the art.
[0094] The Content-to-Topic Matching Module 150 of the system 100
matches accessed content to user-defined topics to produce tagged
content. The architecture and operation of this module will be
described with reference to FIGS. 2 and 3 below.
[0095] The Influence Calculation & Viral Properties Extraction
Module 160 extracts viral properties from the tagged content and
computes influence values based on user-defined influence weights.
This module will be later described with reference to FIGS. 4 and
5. The viral properties or viral dynamics, and the tagged content
are processed by the Recursive Influence Modeling Module 110 to
generate a list of top influencers.
[0096] FIG. 2 shows a block diagram of the Content-to-Topic
Matching Module 150 of the system 100. The diagram 150 shows
entities 220 such as an Individual (Person), a Company or a named
group or organization, which may have one or more outlets where
they publish the content. A content site 210 is defined as a site
hosting one or more outlets, through which an entity or commentor
regularly publishes some form of content or comment. The content
site 210 is accessible to an Internet Crawler 230, which is
connected to a Topic Modeling/Classification Module 240. The Topic
Modeling/Classification Module 240 is also connected to a Topic
Container 250 from which it receives topic-related information. The
Topic Modeling/Classification Module 240 processes the
topic-related information and the content retrieved by the Internet
Crawler 230 to match the content to a defined topic. The matched
content is then stored in a tagged content database 260.
[0097] The Topic Container 250 is a collection of words, phrases,
along with the necessary Boolean logic, which describes a subset of
various social media content, usually centered around a brand,
name, field of study, market, concept, or product. FIG. 2 shows an
exemplary content of the Topic Container 250 presented as a table
identifying a topic profile defined by a Topic_ID 253 and a
collection of keywords and Boolean operators 255 describing the
Topic_ID 253. In this exemplary table, the topic is identified as
"Social media" and the keywords defining the topic include, but are
not limited to, `blogging`, `social media`, `social networking`,
twitter, Facebook, LinkedIn, `video sharing`, etc., which describe
the topic "Social Media", as listed in the keywords and Boolean
operators 255.
[0098] The Topic Modeling/Classification Module 240 defines a topic
model, which is a trained text classification model, created by
feeding a text classifier, a labeled corpus of on-topic and not
on-topic content. The classifier then gauges unlabelled data based
on how closely it matches the trained topic model. Text or content
classification methods are well known to those skilled in the art,
and any suitable classification model can be used to classify and
tag the content.
[0099] The operation of the Content-to-Topic Matching Module 150
will now be described in more detail with reference to a flow chart
300 shown in FIG. 3. At step 310, a user defines a topic container
encapsulating a topic profile 320 against which the retrieved
content needs to be matched. The topic profile 320 is defined by
the set of keywords and Boolean operators 255 and the Topic_ID 253
as shown in FIG. 2.
[0100] At step 340, social media content sites 330 are identified
by crawling the Internet, followed by presenting the discovered
content sites 350 to an analysis phase. All discovered social media
content sites 350 are passed through the analysis phase at step
360, where the content is matched against the topic profile 320. If
the content does not match the topic profile 320 (exit "No" from
step 360), it is disregarded in step 380. If a match is found (exit
"Yes" from step 360), the content is then tagged with the
corresponding topic in step 370.
[0101] FIG. 4 shows a block diagram 400 for the system for
recursively determining the influence of a commentor, an
individual, or a site according to the embodiment of the invention.
A set of Tagged Content 460 is provided in connection with a Viral
Properties Time Series Extractor 450. Each piece of content from
the set of tagged content 460, as described above, is a content
that matches the selected topic profile 320 and is contained or
hosted by a content site. Each piece of content is part of an
entity's outlet such as blog, twitter, streaming video channel,
micromedia or social networking outlet, etc.
[0102] The Viral Properties Time Series Extractor 450 extracts the
viral properties related to the set of tagged content 460 and
stores them in time series so that the history of each viral
property is recorded. Viral properties or viral dynamics are
defined as the various social media popularity metrics. Examples of
viral properties include, but are not limited to: [0103] User
engagement across topically relevant posts, wherein the engagement
is measured by the length of the commenting threads and the number
of unique commentors; [0104] Average Comment count across topically
relevant posts; [0105] Average unique commentor count across
topically relevant posts; [0106] Cited individual count, [0107]
Inbound links across topically relevant posts; [0108] Blog
subscribers across all posts; [0109] Average Social bookmarks
across all topically relevant posts; [0110] Average Social news
votes and buries across all topically relevant posts; [0111] Total
Count of topically relevant posts; and [0112] Total Count of
appearance of Individuals names across all posts.
[0113] Other viral properties include breadth of reply, views,
bookmarks, votes, buries, favorites, awards, acceleration,
momentum, subscription counts, replies, spoofs, ratings, friends,
followers, posts, and updates.
[0114] An Influence Determination Module 440 is connected to the
Viral Properties Time Series Extractor 450 for determining the
influence of each piece of content of the set of tagged content
460. In Determining the influence of a piece of content, the
Influence
[0115] Determination Module 440 receives user-defined influence
weights for each collected viral properties and applies a
mathematical model such as a linear combination model involving
both the weights and the viral properties.
[0116] The Influence Determination Module 440, in this embodiment
is also responsible for the determination of the influence of an
outlet as described in the co-pending U.S. application Ser. No.
12/174,345 filed Jul. 18, 2008 entitled "Method And System For
Determining Topical On-Line Influence Of An Entity" cited
above.
[0117] A Topic-Commentor Influence Modeling Module 410 receives
information related to the names of Unique Commentors and links
from the Viral Properties Time Series Extractor 450 extracted from
the tagged content 460. The links relate to hypertext or other web
links that point back to the outlets such as blog, twitter or
macromedia owned by the unique Commentors. The module 410 creates a
topic-commentor influence model based on the viral properties
across all content where the commentor has a comment, and
determines the topical influence of the commentor characterized by
an influence value of the commentor.
[0118] The module 410 includes also an update module (CI_Updating
Module 411) for updating the commentor influence value. When the
commentor has his own outlet, the CI_Updating Module 411 updates
the influence value of the commentor based on an influence value of
the outlets owned by the commentor and calculated by the Influence
Determination Module 440.
[0119] A Recursive Site Influence Modeling module 420 is provided
to process the set of tagged-content 460 and identify sites hosting
the set of tagged content 460. The module 420 determines the
influence of a site, characterized by a site influence value, from
a linear combination model based on the content influences
generated by the Influence Determination Module 440 by summing the
content influences across all topically relevant content hosted by
the site. Alternatively, the site influence can be computed by
aggregating the extracted viral properties across all topically
relevant content hosted by the site. The aggregation of extracted
viral properties relates to the summing of all individual extracted
viral properties. For example, if Content_1 has two types of viral
properties with the following values: VP_C1_1 and VP_C1_2; and
Content_2 has the same two types of viral properties with the
following values: VP_C2_1 and VP_C2_2; then the aggregated viral
properties would be: (VP_C1_1+VP_C2_1) and (VP_C1_2+VP_C2_2).
[0120] The Recursive Site Influence Modeling module 420 includes a
SI_Updating module 412 for recursively fine-tuning the site
influence model by including the influence value of the commentor
from the topic-commentor influence model as well as an influence
value for an individual calculated from a topic-individual
influence model. In this fine-tuning operation, the site-influence
model uses the influence of commentors commenting on the site and
the influence of individuals named on the site to recursively
update the model and determine the site influence. This process
will be described hereinafter with reference to FIG. 5.
[0121] A Topic-Individual Influence Modeling Module 430 identifies
individuals named in the content and creates the topic-individual
influence model based on an aggregation of viral dynamics of
content citing the individual. The Topic-Individual Influence
Modeling Module 430 further includes an I_Updating module 413 for
updating the influence of the individual characterized by the
influence value of the individual using respective influence values
of the outlets owned by the individual, when available. The
operation of the Topic-Individual Influence Modeling Module 430
will be described in more details in reference to FIG. 5.
[0122] FIG. 5 shows a flowchart 500 illustrating the operation of
the system 400 of FIG. 4.
[0123] In the embodiment of the invention, each piece of content
505 that matches the topic profile 320 is scheduled to have its
viral properties extracted at step 510 on a regular schedule, and
stored in time series so that the history of each viral property is
recorded. The time schedule used for extracting the viral
properties changes as the recorded viral property values are
analyzed. For instance, if upon checking the viral properties for a
blog post on a 3 hour schedule, it is determined that the number of
new comments has exploded, then, the schedule will be altered to
ensure that the viral properties are checked more frequently.
Conversely, if the comment count has changed little or not at all,
the schedule may be changed to half the frequency, down to every 6
hours. Conveniently, different viral properties may have same or
different time extraction schedule.
[0124] The extracted viral properties are used at "Determine
Influence Based on Viral Properties of Content", step 520, to
determine the influence of each piece of content 505 based on the
collected viral properties. At step 520, the influence values of
outlets are also calculated as disclosed in the above-mentioned
patent application cited above.
[0125] As an example, the influence value for the piece of content
505 can be calculated according to a linear combination model of
the viral properties weighted with respective weights applied to
each of the viral properties:
CalculatedContentInfluence=(Weight1*VP1)+(Weight2*VP2)+(Weight3*VP3)+(We-
ight4*VP4)+(Weight5*VP5)+(Weight6*VP6)++(Weightn*VPn),
where Weight1-Weightn are respective weight factors defining the
relevant contribution of various viral properties (VP1-VPn) into
the topical on-line influence for this content, and the topical
on-line influence is represented by a value conveniently normalized
to a scale of 0-100.
[0126] In the embodiment of the present invention, a user is
responsible for adjusting the weights (i.e. Weight1-Weightn) for
various viral properties used in the model above to reflect the
viral properties which, in user's opinion, are most suitable for
the business goals the user or his clients have set forth. The
weight for each type of the outlet is also user defined and is
entered at "User-Defined Influence Calculation Weights" step
525.
[0127] It is assumed that same default weights can be applicable to
an average user. The weights adjusted by a user are saved on a per
topic basis, allowing for a different topic to have a different
weighting system to align with potentially different business
goals.
[0128] At "Extract Unique Commentor Names and Links", step 530,
names of commentors commenting on the piece of content 505 are
extracted along with any link back to the outlets owned by the
commentor. As an example, Robert Scobble may be a commentor on the
piece of content 505, and he may also be providing links to his own
outlets such as blog, twitter site, Facebook profile, and streaming
video channel (Kyte.tv/RobertScobble).
[0129] At step 540, the Commentor Influence model is created from
the aggregation of the viral properties across all content having a
comment from the selected commentor.
[0130] The influence value computed from Commentor Influence model,
termed herein as Commentor_AggregatedViralsInfluence, is calculated
according to a linear combination model of the extracted viral
properties weighted with respective weights applied to each of the
extracted viral properties as shown in the following model:
Commentor_AggregatedViralsInfluence=(Weight_11*AVP1)+(Weight_22*AVP2)+(W-
eight_33*AVP3)+ . . . (Weight_nn*AVPn),
where Weight_11-Weight_nn are respective weight factors defining
relevant contribution of various viral properties into the modeling
of the Commentor_AggregatedViralsInfluence, and where AVP1-AVPn are
respective Aggregated Viral Properties extracted across all pieces
of content 505 containing a comment from the selected commentor,
namely Robert Scobble in this example.
[0131] The Commentor Influence model, according to the embodiment
of the present invention, is further updated by the CI_Updating
Module 411 based on the influence values of all outlets owned by
the commentor Robert Scobble.
[0132] Considering that the commentor, Robert Scobble, has his own
outlets, namely a blog, video channel, MicroMedia channel, social
network and social news page and image sharing page, The Commentor
Influence model can then be updated according to a linear
combination model of the topical influence values of the outlets
owned by Robert Scobble and the topical influence value of the
commentor calculated above as Commentor_AggregatedViralsInfluence
as shown in the following model:
CommentorInfluence=((Weight000*Commentor_AggregatedViralsInfluence)+(Wei-
ght111*Commentor_CalculatedBlogInfluence)+(Weight222*Commentor_Calculated
VideoChannelInfluence)+(Weight333*Commentor_CalculatedMicroMediaInfluence-
)+(Weight444*Commentor_CalculatedSocialNewsInfluence)+(Weight555*Commentor-
_CalculatedSocialNetworkInfluence)+(Weight666*Commentor_CaculatedImageShar-
ingInfluence))/7.
where Weight111-Weight666 are respective weight factors assigned to
the various forms of outlets, Weight000 is the weight factor for
the Commentor_AggregatedViralsInfluence value into the final
Commentor influence value, and where
Commentor_CalculatedBlogInfluence,
Commentor_CalculatedVideoChannelInfluence,
Commentor_CalculatedMicroMediaInfluence,
Commentor_CalculatedSocialNewsInfluence,
Commentor_CalculatedSocialNetworkInfluence, and
Commentor_CaculatedImageSharingInfluence
[0133] are respective influence values of the various outlets owned
by Robert Scobble calculated at step 520 as disclosed in the
above-mentioned patent application incorporated herein by
reference.
[0134] An influence value characterizing the
Commentor_CalculatedBlogInfluence can be calculated using a linear
combination model as follows:
Commentor_CalculatedBlogInfluence=(Weight1*BlogEngagement)+(Weight2*Aver-
age comment Count)+(Weight3*Average Unique Commentor
Count)+(Weight4*Inbound Links)+(Weight5*Blog
Subscribers)+(Weight6*Bookmarks)+(Weight7*Votes)+(Weight8*Count of
Topically relevant posts),
where Weight1-Weight8 are respective weight factors defining the
relevant contribution of various viral properties into the topical
on-line influence for Robert Scobble blog.
[0135] Likewise the influence value characterizing the
Commentor_CalculatedVideoChannelInfluence can be calculated as
follows:
Commentor_CalculatedVideoChannelInfluence=(Weight11*Average
Concurrent Viewership)+(Weight12*Total Views)+(Weight13*Inbound
Links)+(Weight14*Engagement)+(Weight15*Average Comment
Count)+(Weight16*Unique Commentor Count)+(Weight17*Count of
Topically Relevant Posts)
where Weight11-Weight17 are respective weight factors defining the
relevant contribution of various viral properties into the topical
on-line influence of this video channel outlet.
[0136] Using the linear combination models above, the influence
value characterizing the topical on-line influence of all the
outlets can be calculated by taking into account specific viral
properties that are measurable in the selected outlet.
[0137] As each viral property value is stored in time series, the
topical on-line influence of all commentors can be determined for
any time or time periods, and top influential commentors can be
identified or surfaced at step 545.
[0138] In one embodiment of the invention, topical on-line
influences for individuals are also determined. As shown in the
flowchart 500, at step 550, names of individuals cited in all
pieces of content 505 are extracted, and their viral properties are
aggregated around each individual to create a topic-Individual
influence model (step 560). As an example, an individual named Cesc
may be cited in one or more pieces of content 505. In this example,
the topical on-line influence of Cesc can be determined by
calculating the influence value of the individual, (i.e. Cesc)
based on a linear combination model. The model can be based on
aggregated viral dynamics of all the content where Cesc is cited,
and applying user-defined weights to the aggregated viral dynamics,
as follows:
Individual_AggregatedViralsInfluence=(Weight_01*AVD_01)+(Weight_02*AVD_0-
2)+(Weight_03*AVD-03)+ . . . (Weight_0n*AVD_0n)
where Weight_01-Weight_0n are respective weight factors defining
relevant contribution of various viral dynamics into the influence
of the individual, and where AVD_01-AVD_0n are respective
Aggregated Viral Dynamics extracted across all pieces of content
505 mentioning the selected individual (i.e. Cesc).
[0139] Furthermore, for individuals who have their own social media
outlet such as blog, video channel or other social media page, the
topic-individual influence model can integrate the influence of
those outlets using the I_Updating module 413 of the
Topic-Individual Influence Modeling Module 430 according to the
following linear combination model:
IndividualInfluence=((Weight_000*Individual_AggregatedViralsInfluence)+(-
Weight_111*Individual_CalculatedBlogInfluence)+(Weight_222*Individual_Calc-
ulatedVideoChannelInfluence)+(Weight_333*Individual_CalculatedMicroMediaIn-
fluence)+(Weight_444*Individual_CalculatedSocialNewsInfluence)+(Weight_555-
*Individual_CalculatedSocialNetworkInfluence)+(Weight_666*Individual_Cacul-
atedlmageSharingInfluence))/7.
[0140] where Weight_111-Weight_666 are respective weight factors
defining relevant contribution of various forms of outlets, and
Weight_000 is the weight factor for the
Individual_AggregatedViralsInfluence value into the final
Individual influence value.
[0141] Using this method, all individuals can be assigned an
associated influence value, followed by identifying top individual
influencers at step 570.
[0142] At step 580, a site influence model referred also as
topic-site influence model is recursively created for each content
site hosting one or more content matching the selected topic
profile based on the aggregation of the viral properties for all
pieces of content 505 within each site and an influence weight
(Site_type_weight) according to the type of outlet hosted by the
site (e.g. blog, twitter, social media site, etc). When a content
site includes content, in which an influential commentor is
present, the SI_updating module 412 fine-tunes the site influence
model by recursively applying the influence value of the commentor
in the site influence model. In this embodiment, the influence
value of the commentor is added to the site influence model in a
weighted manner. The influence value of a site named S1 is
calculated according to the following linear combination model:
CalculatedSiteInfluence_S1=Site_type_weight*(W0*S1_AVP0+ . . .
Wn*S1_AVPn)+(CW0*CommentorInfluence_0+ . . .
CWk*Commentorinfluence_k)
where Site_type_weight is the weight defined for the type of site
(blog site, video channel provider site, twitter type of site,
etc); W0 . . . Wn are respective weights for the aggregate viral
properties S1_AVP0 . . . S1_AVPn collected on all content of the
site S1, and CW0 . . . CWk are respective weights applied to the
influence values of the commentors CommentorInfluence_0 . . .
CWn*CommentorInfluence_k. The weights CWi as well as the number of
commentors can be user-defined and set according to business
objectives, or a threshold of influence. Other criteria for setting
these values can as well be adopted.
[0143] The above noted model is exemplary, it is contemplated that
other models can as well be used to determine the influence value
of a site.
[0144] According to another embodiment, one or more top individual
influencers can be included in the site influence model described
with reference to step 580. Thus, a modified model for calculating
the influence value of the site becomes as follows:
CalculatedSiteInfluence_S1=Site_type_weight*(W0*S1_AVP0+ . . .
Wn*S1_AVPn)+(CW0*CommentorInfluence_0+ . . .
CWk*Commentorinfluence_k)+(IW0*IndividualInfluence_0+ . . .
IW1*IndividualInfluence_1)
where IW0 . . . IW1 are respective weights applied to the influence
values of individuals IndividualInfluence_ . . .
IndividualInfluence_1.
[0145] By iterating the determination of the site influence
characterized by the site influence value across all sites having
content that matches the topic container, top Site Influencers can
be identified (step 585).
[0146] The embodiments of the present invention provide numerous
advantages, most importantly, to public relation professionals to
make preemptive marketing decisions that are not available today.
Within a selected topic, the present invention allows marketers to
compare two different sites, each hosting a set of outlets, and
determine which one has the most topical on-line influence, and
therefore has more potential to induce a successful advertising
campaign.
[0147] As mentioned above, the system of the present invention have
been implemented as a general purpose or specialized computer
having a CPU and a memory storing computer code instructions to be
performed by the CPU, which form the modules of the system perform
as shown in FIGS. 1 and 4 described above, and, when executed,
perform the steps of the method described above.
[0148] A computer readable medium is also provided, e.g., CD-ROM,
DVD, flash memory, floppy or the like, wherein the computer code
instructions are stored, for executing by a computer to perform the
steps of the method described above.
[0149] Thus, improved methods and system for determining topical
on-line influence of a site, commentor, and an individual have been
provided.
* * * * *
References