Automated media analysis and document management system

• US 7,860,872 B2
• Filed: 01/29/2007
• Issued: 12/28/2010
• Est. Priority Date: 01/29/2007
• Status: Active Grant

• Current Assignee: Agility PR Solutions Canada Ltd.
• Original Assignee: Nikip Technology Limited
• Inventors: Serjeantson, Brett, Williamson, Paul
• Primary Examiner(s): Trujillo; James
• Assistant Examiner(s): PHILLIPS, III, ALBERT M
• Application Number:
  US11/668,346
• Publication Number:
  US 20080183710A1
• Time in Patent Office: 1,429 Days
• Field of Search: 707/917, 707/755, 707/811
• US Class Current: 707/755
• CPC Class Codes: G06F 16/30 of unstructured textual data document management systems G06F16/93
• CPC Class Codes: G06F 2216/01 Automatic library building
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: Y10S 707/917 Text

Abstract

A web-based media analysis system, consisting of automated media analysis and document management tools, which processes news articles by parsing the news contents or documents and assigning, relating, and extracting information from the news contents for media analysis and relationally storing them in at least one database. The system further comprises a toning engine for toning articles accurately, based on words, attributes and categories of the article, and optionally based on the author of the article, if applicable.

Claims

1. A web-based media analysis computer system for analyzing at least one media content that includes text, comprising:

i. a memory comprising at least one database being accessible to the computer system;

ii. the computer system implementing an uploader that uploads said at least one media content from at least one content provider over a communication network;

iii. the computer system implementing a parser that converts each of said at least one media content into serialized data, filters out unessential data from said serialized data according to a predefined list of unessential data, rationalizes nouns, pronouns and names in said each of said at least one media content, and extracts attributes and categories of said each of said at least one media content, one or more quotes, and attributes of said one or more quotes from said serialized and filtered data of said each of said at least one media content using regular expressions and predefined parsing rules, relationally stores and cross-references said serialized and filtered data, said extracted attributes and said categories of said each of said at least one media content, said one or more quotes, and said attributes of said one or more quotes into a plurality of tables in said at least one database, wherein said attributes of said one or more quotes comprise the name of the quoted person or organization; and

iv. the computer system implementing an analysis module that retrieves data from said plurality of said tables in said at least one database and wherein said analysis module further comprises a toning engine that determines a tone level of said each of said at least one media, said toning engine is provided with tone level probabilities of meaningful minimum sections, tone level probabilities for attributes, and tone level probabilities for categories in said at least one database, said tone engine;

a. parses each of said at least one media content and splitting into serialized meaningful minimum sections;

b. retrieves, from said at least one database, said tone level probabilities for each of said serialized meaningful minimum sections that cause said each of said at least one media content to be toned at, and determines most probable tone levels of said each of said at least one media content based on said tone level probabilities of said serialized meaningful minimum sections;

c. retrieves, from said at least one database, said extracted attributes of said each of said at least one media content and said tone level probabilities for each of said attributes of said each of said at least one media content, and determines most probable tone levels of said each of said at least one media content based on said tone level probabilities of said attributes;

d. retrieves, from said at least one database, said extracted categories of said each of said at least one media content and said tone level probabilities for each of said categories of said each of said at least one media content, and determines most probable tone levels of said each of said at least one media content based on said tone level probabilities of said categories; and

establishes said most probable tone level of said each of said at least one media content by weighing and ranking said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said each of said serialized meaningful minimum sections, said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said attributes, and said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said categories; and

e. generates a report in response to a request from a client browser.

2. The web-based media analysis system recited in claim 1, wherein said meaningful minimum sections are words.

3. The web-based media analysis system recited in claim 1, wherein said toning engine further weighs tones of authors of said at least one media content for establishing said most probable tone level of said each of said at least one media content, wherein said tones of said authors of said at least one media content are established by retrieving, from said at least one database, tone level probabilities for the author of said each of said at least one media content, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said authors.

4. The web-based media analysis system recited in claim 3, wherein said authors are individuals or organizations.

5. The web-based media analysis system recited in claim 1, wherein said tone level probabilities for all of said serialized meaningful minimum sections, said tone level probabilities for said attributes of said each of said at least one media content, and said tone level probabilities for said categories of said each of said at least one media content are determined through training said toning engine by feeding a plurality of sample media contents with suggested tones thereof and updating said at least one database.

6. The web-based media analysis system recited in claim 5, wherein said training comprises the steps of:

i. preparing said plurality of sample media contents with suggested tones thereof;

ii. splitting each of said plurality of said sample media contents into meaningful minimum sections while filtering out any unessential sections and extracting attributes and categories of said each of said plurality of said sample media contents using regular expressions;

iii. tracking number of appearances of each of said meaningful minimum sections in said each of said plurality of said sample media contents and associating said number of appearances and said suggested tone corresponding to said each of said plurality of said sample media contents, deriving said tone level probabilities of said each of said meaningful minimum sections cause said each of said plurality of said sample media contents to be toned at, and updating said at least one database with said tone level probabilities of said meaningful minimum sections;

iv. deriving said tone level probabilities of said attributes of said each of said plurality of said sample media contents cause said each of said plurality of said sample media contents to be toned at, and updating said at least one database with said tone level probabilities of said attributes; and

v. deriving said tone level probabilities of said categories of said each of said plurality of said sample media contents that cause said each of said plurality of said sample media contents to be toned at, and updating said at least one database with said tone level probabilities of said categories.

7. The web-based media analysis system recited in claim 6, wherein said meaningful minimum sections are words.

8. A computer-implemented method for managing and analyzing at least one media content comprising the steps of:

i. uploading said at least one media content from at least one content provider;

ii. converting each of said at least one media content into serialized data, filtering out unessential data from said serialized data according to a predefined list of unessential data, rationalizing nouns, pronouns and names therein, extracting attributes and categories of said each of said at least one media content, one or more quotes, and attributes of said one or more quotes in said each of said at least one media content using regular expressions, wherein said attributes of said one or more quotes comprise the name of the quoted person or organization;

iii. relationally storing and cross-referencing said serialized and filtered data and said extracted attributes and said categories of said each of said at least one media content, said one or more quotes, and said attributes of said one or more quotes into a plurality of tables in at least one database; and

iv. retrieving data from said plurality of tables in said at least one database in response to a request from a client browser and generating a report and toning said at least one media content prior to said generating said report by carrying out the process comprising the steps of;

a. parsing each of said at least one media content and splitting into serialized meaningful minimum sections;

b. retrieving, from said at least one database, tone level probabilities for each of said serialized meaningful minimum sections that cause said each of said at least one media content to be toned, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said serialized meaningful minimum sections;

c. retrieving, from said at least one database, tone level probabilities for each of said attributes of said each of said at least one media content, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said attributes;

d. retrieving, from said at least one database, tone level probabilities for each of said categories of said each of said at least one media content, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said categories; and

e. establishing said most probable tone level of said each of at least one media content by weighing and ranking said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said each of said serialized meaningful minimum sections, said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said attributes, and said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said categories.

9. The method for managing and analyzing said at least one media content recited in claim 8, wherein said meaningful minimum sections are words.

10. The method for managing and analyzing media contents recited in claim 8, wherein said toning engine further weighs tones of authors of said at least one media content for establishing said most probable tone level of said each of said at least one media content, wherein said tones of said authors of said at least one media content are established by retrieving, from said at least one database, tone level probabilities for the author of said each of said at least one media content, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said authors.

11. The method for managing and analyzing said at least one media content recited in claim 10, wherein said authors are individuals or organizations.

12. The method for managing and analyzing said at least one media content recited in claim 8, wherein said tone level probabilities for all of said serialized meaningful minimum sections, said tone level probabilities for said attributes of said each of said at least one media content, and said tone level probabilities for said categories of said each of said at least one media content are determined through a training process by feeding a plurality of sample media contents with suggested tones thereof and updating said at least one database, wherein said training process comprises the steps of:

i. splitting each of said plurality of said sample media contents into meaningful minimum sections while stripping out any unessential sections and extracting attributes and categories of said each of said plurality of said sample media contents using regular expressions;

ii. tracking number of appearances of each of said meaningful minimum sections in said each of said plurality of said sample media contents and associating said number of appearances and said suggested tone corresponding to said each of said plurality of said sample media contents, deriving said tone level probabilities of said each of said meaningful minimum sections cause said each of said plurality of said sample media contents to be toned at, and updating said at least one database with said tone level probabilities of said meaningful minimum sections;

iii. deriving said tone level probabilities of said attributes of said each of said plurality of said sample media contents cause said each of said plurality of said sample media contents to be toned at, and updating said at least one database with said tone level probabilities of said attributes; and

iv. deriving said tone level probabilities of said categories of said each of said plurality of said sample media contents cause said each of said plurality of said sample media contents to be toned at, and updating said at least one database with said tone level probabilities of said categories.

13. The method for managing and analyzing said at least one media content recited in claim 12, wherein said meaningful minimum sections are words.

14. A tangible computer readable medium storing executable computer program instructions which, when executed at a server, cause the server to perform a process for analyzing at least one media content, the process comprising the steps of:

i. unloading said at least one media content from at least one content provider over a communication network;

ii. converting each of said at least one media content into serialized data, filtering out unessential data from said serialized data according to a predefined list of unessential data, extracting attributes and categories of said each of said at least one media content, and extracting, one or more quotes and attributes of said one or more quotes in said each of said at least one media content using regular expressions, wherein said attributes of said one or more quotes comprise the name of the quoted person or organization;

iii. relationally storing and cross-referencing said serialized and filtered data, said extracted attributes, said extracted categories and said one or more quotes into a plurality of tables in at least one database;

iv. retrieving data from said plurality of said tables in said at least one database and generating a report in response to request from a user; and

v. toning said at least one media content prior to said generating said report by carrying out the process steps of;

a. parsing each of said at least one media content and splitting into serialized meaningful minimum sections;

b. retrieving, from said at least one database, tone level probabilities for each of said serialized meaningful minimum sections that cause said each of said at least one media content to be toned at, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said serialized meaningful minimum sections;

c. retrieving, from said at least one database, tone level probabilities for each of said attributes of said media content, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said attributes;

d. retrieving, from said at least one database, tone level probabilities for each of said categories of said each of said at least one media content, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said categories; and

e. establishing said most probable tone level of said each of said at least one media content by weighing and ranking said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said each of said serialized meaningful minimum sections, said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said attributes, and said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said categories.

15. The computer readable medium recited in claim 14, wherein said meaningful minimum sections are words.

16. The computer readable medium recited in claim 15, wherein said toning engine further weighs tones of authors of said at least one media content for establishing said most probable tone level of said each of said at least one media content, wherein said tones of said authors of said at least one media content are established by retrieving, from said at least one database, tone level probabilities for the author of said each of said at least one media content, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said authors.

17. The computer readable medium recited in claim 16, wherein said authors are individuals or organizations.

18. The computer readable medium recited in claim 14, wherein said tone level probabilities for all of said serialized meaningful minimum sections, said tone level probabilities for said attributes of said each of said at least one media content, and said tone level probabilities for said categories of said each of said at least one media content are determined through a training process by feeding a plurality of sample media contents with suggested tones thereof and updating said at least one database, wherein said training process comprises the steps of:

i. splitting each of said plurality of said sample media contents into meaningful minimum sections while stripping out any unessential sections and extracting attributes and categories of said each of said plurality of said sample media contents using regular expressions;

ii. tracking number of appearances of each of said meaningful minimum sections in said each of said plurality of said sample media contents and associating said number of appearances and said suggested tone corresponding to said each of said plurality of said sample media contents, *deriving said tone level probabilities of said each of said meaningful minimum sections, and updating said at least one database with said tone level probabilities of said meaningful minimum sections;

iii. deriving said tone level probabilities of said attributes of said each of said plurality of said sample media contents cause said each of said plurality of said sample media contents to be toned at, and updating said at least one database with said tone level probabilities of said attributes; and

iv. deriving said tone level probabilities of said categories of said each of said plurality of said sample media contents cause said each of said plurality of said sample media contents to be toned at, and updating said at least one database with said tone level probabilities of said categories.

19. The computer readable medium recited in claim 18, wherein said meaningful minimum sections are words.

20. A computer-implemented method of toning at least one media contents comprising the steps of:

i. parsing each of said at least one media content, splitting and serializing into meaningful minimum sections, and extracting attributes and categories of said each of said at least one media content, one or more quotes, and attributes of said one or more quotes in said each of said at least one media content;

ii. retrieving, from at least one database, tone level probabilities for each of said serialized meaningful minimum sections that cause said each of said at least one media content to be toned at, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said serialized meaningful minimum sections;

iii. retrieving, from said at least one database, tone level probabilities for each of said attributes of said each of said at least one media content, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said attributes;

iv. retrieving, from said at least one database, tone level probabilities for each of said categories of said each of said at least one media content, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said categories; and

v. establishing said most probable tone level of said each of said at least one media content by weighing and ranking said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said each of said serialized meaningful minimum sections, said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said attributes, and said most probable tone levels of said each of said at least one media content based on said tone level probabilities of said categories.

21. The method of toning said at least one media content recited in claim 20, wherein said meaningful minimum sections are words.

22. The method of toning said at least one media content recited in claim 20, wherein said toning engine further weighs tones of authors of said at least one media content for establishing said most probable tone level of said each of said at least one media content, wherein said tones of said authors of said at least one media content are established by looking up from said at least one database probabilities of tone levels for the author of said each of said at least one media content, and determining most probable tone levels of said each of said at least one media content based on said tone level probabilities of said authors.

23. The method of toning said at least one media content recited in claim 22, wherein said authors are individuals or organizations.

24. The method of toning said at least one media content recited in claim 20, wherein said tone level probabilities for all of said serialized meaningful minimum sections, said tone level probabilities for said attributes of said each of said at least one media content, and said tone level probabilities for said categories of said each of said at least one media content are determined through a training process by feeding a plurality of sample media contents with suggested tones thereof and updating said at least one database, wherein said training process comprises the steps of:

i. splitting each of said plurality of said sample media contents into meaningful minimum sections while stripping out any unessential sections and extracting attributes and categories of said each of said plurality of said sample media contents using regular expressions;

ii. tracking number of appearances of each of said meaningful minimum sections in said each of said plurality of said sample media contents and associating said number of appearances and said suggested tone corresponding to said each of said plurality of said sample media contents for deriving probabilities of said each of said meaningful minimum sections cause said each of said plurality of said sample media contents to be toned at, and updating said at least one database with said tone level probabilities of said meaningful minimum sections;

iii. deriving said tone level probabilities of said attributes of said each of said plurality of said sample media contents, and updating said at least one database with said tone level probabilities of said attributes; and

iv. deriving said tone level probabilities of said categories of said each of said plurality of said sample media contents, and updating said at least one database with said tone level probabilities of said categories.

25. The method of toning said at least one media content recited in claim 24, wherein said meaningful minimum sections are words.