Method of monitoring electronic media

• US 8,010,524 B2
• Filed: 10/29/2007
• Issued: 08/30/2011
• Est. Priority Date: 10/29/2007
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Chen, Ying, Griffin, Thomas D., Proctor, Larry L., Spangler, W. Scott, Behal, Amit
• Primary Examiner(s): Ali; Mohammad
• Assistant Examiner(s): HASAN, SYED HAROON
• Application Number:
  US11/927,238
• Publication Number:
  US 20090119275A1
• Time in Patent Office: 1,401 Days
• Field of Search: None
• US Class Current: 707/722
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

Consumer-generated media (CGM) and/or other media are monitored to allow an organization to become aware of, and respond to, issues that may affect how it is perceived by the public. An extract, transform, load (ETL) engine is used to process CGM and other media content, and an analytical engine utilizes a multi-step progressive filtering approach to identify those documents that are most relevant. The filtering approach includes executing broad queries to extract relevant content from different CGM and other sources, extracting text snippets from the relevant content and performing de-duplication, defining organizational identity (e.g., brand name, trade name, or company name) and hot-topic models using a rule-based and statistical-based approach, and using the models together in an orthogonal filtering approach to effectively generate alerts and reports. The methodology is found to be substantially more effective compared to a conventional keyword based approach.

Claims

1. A computer-implemented method, comprising:

receiving at least one keyword-based query from a user device;

applying said at least one keyword-based query to a collection of documents to determine which ones of them are of potential interest, the documents of potential interest thereby forming a subset of the collection of documents;

extracting text snippets from documents in the subset, each of the extracted text snippets including at least one term of interest;

applying, to text snippets extracted from documents in the subset, a name-based text pattern model, and applying, to text snippets extracted from documents in the subset, a topic-based text pattern model that is independent from the name-based text pattern model, thereby identifying text snippets having references to both i) a predetermined entity of interest known by a given name and ii) a predetermined topic of interest; and

identifying documents corresponding to the identified text snippets, wherein;

i) the name-based text pattern model excludes from consideration text snippets in which the given name appears but which in context does not refer to the entity of interest,

ii) the topic-based text pattern model specifies at least one unwanted text pattern that matches a topic other than the predetermined topic of interest, so that text snippets including said at least one unwanted text pattern are excluded from consideration,

iii) each of the models uses at least one regular expression, and

iv) the given name is selected from the group consisting of trade names, brand names, and company names.

2. The method of claim 1, further comprising periodically providing output to an interested party.

3. The method of claim 2, comprising:

applying the name-based text pattern model to extracted text snippets to determine which ones of them have a reference to the predetermined entity of interest, thereby forming a subset of the extracted text snippets; and

applying the topic-based text pattern model to text snippets in the subset of the extracted text snippets, thereby identifying text snippets having references to both i) the predetermined entity of interest and ii) the predetermined topic of interest, wherein each of the text snippets is no longer than 50 words.

4. The method of claim 1, comprising identifying and eliminating any duplicate among the extracted text snippets, thereby forming a set of remaining text snippets.

5. The method of claim 1, further comprising:

reading at least portions of said identified documents; and

identifying potential newsworthy trends in view of said reading.

6. The method of claim 1, wherein the given name is a trade name.

7. The method of claim 1, wherein the given name is a name of a company.

8. The method of claim 1 wherein the given name is a brand name.

9. The method of claim 1 wherein each of the text snippets is no longer than three sentences.

10. The method of claim 1, further comprising sorting the identified documents according to a measure of their relevance.

11. The method of claim 1 wherein the collection of documents includes documents taken from consumer-generated media.

12. At least one tangible computer-useable storage medium, said at least one medium having a computer-readable program, wherein the program upon being processed on a computer causes the computer to implement the steps of:

applying at least one keyword-based query to a collection of documents to determine which ones of them are of potential interest, the documents of potential interest thereby forming a subset of the collection of documents;

extracting text snippets from documents in the subset, each of the extracted text snippets including at least one term of interest;

applying, to text snippets extracted from documents in the subset, a name-based text pattern model, and applying, to text snippets extracted from documents in the subset, a topic-based text pattern model that is independent from the name-based text pattern model, thereby identifying text snippets having references to both i) a predetermined entity of interest known by a given name and ii) a predetermined topic of interest; and

identifying documents corresponding to the identified text snippets, wherein;

i) the name-based text pattern model excludes from consideration text snippets in which the given name appears but which in context does not refer to the entity of interest,

ii) the topic-based text pattern model specifies at least one unwanted text pattern that matches a topic other than the predetermined topic of interest, so that text snippets including said at least one unwanted text pattern are excluded from consideration,

iii) each of the models uses at least one regular expression, and

iv) the given name is selected from the group consisting of trade names, brand names, and company names.