Framework for generating sentiment data for electronic content

• US 8,396,820 B1
• Filed: 04/28/2010
• Issued: 03/12/2013
• Est. Priority Date: 04/28/2010
• Status: Active Grant

• Current Assignee: Douglas Rennie
• Original Assignee: Douglas Rennie
• Inventors: Rennie, Douglas
• Primary Examiner(s): Gaffin, Jeffrey A
• Assistant Examiner(s): Chubb, Mikayla
• Application Number:
  US12/769,315
• Time in Patent Office: 1,049 Days
• Field of Search: None
• US Class Current: 706/47
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique

Abstract

Methods, systems, and techniques for generating sentiment data are provided. Example embodiments provide a Sentiment Thesaurus Framework (STF), which provides a sentiment vocabulary meeting a set of criteria and a sentiment thesaurus according to certain connectedness rules. In one embodiment, the STF provides the sentiment thesaurus by relating connected adjective word-senses to valid sentiment adjective word-senses. The degree of connectedness is limited to one shift in meaning from one adjective to another, which implies at most two non-satellite or axle adjectives between a valid sentiment and a related adjective word-sense.

Claims

1. A non-transitory computer-readable storage medium containing structured data for use with computer program code executing on a computer processor, comprising:

a sentiment thesaurus data structure stored in the storage medium and including data used by the computer program code, when executed, to generate one or more valid sentiments related to an indicated input word, the valid sentiments selected based upon an evaluation of a set of rules that define valid sentiments, the data structure comprising;

a plurality of sentiment entries, each entry containing an indication of one of a plurality of valid sentiment adjectives that can be used to express one of the valid sentiments applicable to online content, the valid sentiments selected based upon the evaluation of the set of rules that define valid sentiments, an indication of an adjective related to the valid sentiment adjective of the sentiment entry, and a connection type indicator that characterizes a semantic relationship between the related adjective and the valid sentiment adjective that is within a specified permissible degree of distance in meaning, the indicated connection type being one of a plurality of permitted connection types that limit a degree of distance in meaning between adjectives.

2. The computer-readable storage medium of claim 1 wherein the computer program code, when executed, generates from the sentiment thesaurus data structure one or more valid sentiments related to an indicated input word by looking up each sentiment entry in which the indicated input word is stored as a related adjective and retrieving an indication to the related valid sentiment adjective.

3. The computer-readable storage medium of claim 2 wherein the computer program code, when executed, returns related valid sentiment adjectives in an order corresponding to the indicated connection type associated with each sentiment entry of each related valid sentiment adjective.

4. The computer-readable storage medium of claim 1 wherein the plurality of permitted connection types include semantic relationships that specify a maximum of one shift in meaning between words.

5. The computer-readable storage medium of claim 1 wherein the plurality of permitted connection types include semantic relationships that specify a first and a second word having a similar meaning, where the first word is a more specific description of the second word.

6. The computer-readable storage medium of claim 1 wherein the plurality of permitted connection types include semantic relationships that specify that two adjectives are connected in a “

similar to”

relationship and/or semantic relationships that specify that two adjectives relate to each other through at most two “

also see”

relationships.

7. The computer-readable storage medium of claim 1 wherein the plurality of permitted connection types include semantic relationships that specify that two adjectives are connected via a maximum of two non-satellite adjectives.

8. The computer-readable storage medium of claim 1 wherein the plurality of permitted connection types include semantic relationships that represent one shift in meaning from a related adjective to a related valid sentiment adjective.

9. The computer-readable storage medium of claim 1 wherein the plurality of permitted connection types include semantic relationships that represent that a related adjective has a meaning similar to a related valid sentiment adjective.

10. The computer-readable storage medium of claim 1 wherein the plurality of permitted connection types comprise semantic relationships that specify that a related adjective can relate to a valid sentiment adjective as one or more of matching, as sharing a valid sense, as sharing a spoke-axle connection to a valid sense, as sharing an axle-axle connection to a valid sense, as sharing a spoke-axle-spoke connection to a valid sense, as sharing an axle-axle-spoke connection to a valid sense, as sharing a spoke-axle-axle connection to a valid sense, or as sharing a spoke-axle-axle-spoke connection to a valid sense.

11. The computer-readable storage medium of claim 1 wherein one or more of the plurality of sentiment entries include a related adjective that is not a valid sentiment, such that the thesaurus data structure may be used by the computer program code to retrieve one or more related valid sentiment adjectives in response to an indicated input word that is not a valid sentiment.

12. The computer-readable storage medium of claim 1 wherein the plurality of sentiment entries correspond to an electronically stored data repository that stores a vocabulary of sentiment adjectives produced by analyzing a universe of adjectives according to a set of rules that determine whether each adjective may be usable as a valid expression of sentiment for online content.

13. The computer-readable storage medium of claim 12 wherein the set of rules employed to determine whether each sense of an adjective may be usable as a valid expression of sentiment for online content comprises:

one or more of;

a rule that excludes a sense of an adjective from the set if the adjective does not have a meaning that can describe an individual piece of content;

a rule that excludes a sense of an adjective from the set if the adjective cannot describe a single noun in the online content without context provided by additional words of the online content;

a rule that excludes a sense of an adjective from the set if the adjective is a number;

a rule that excludes a sense of an adjective from the set if the adjective describes content from a third person perspective;

a rule that excludes a sense of an adjective from the set if the adjective is limited to the predicative or adjunctive state in common usage of the adjective;

a rule that excludes a sense of an adjective from the set if the adjective describes content non-subjectively in a manner that could describe all content;

a rule that excludes a sense of an adjective from the set if the adjective describes content as if existing as a tangible thing;

a rule that excludes a sense of an adjective from the set if the adjective describes content as related to a world-at-large;

a rule that excludes a sense of an adjective from the set if the adjective describes content in a sense of categorization;

a rule that excludes a sense of an adjective from the set if the adjective describes content in regard to formal structure of a content as a medium;

a rule that excludes a sense of an adjective from the set if the adjective is not a single word and is a lexical unit comprised of multiple aggregated words;

a rule that excludes a sense of an adjective from the set if the adjective describes content as undergoing or having undergone change;

a rule that excludes a sense of an adjective from the set if the adjective is an uncommon usage spelling variation of another adjective;

a rule that excludes a sense of an adjective from the set if the adjective has a negative prefix or suffix and has a meaning that is synonymous with an adjective already included in the vocabulary; and

/or

a rule that excludes a sense of an adjective from the set if the adjective anthropomorphizes corresponding content as a proxy for the author in a manner that fails to describe tonality of the corresponding content.

14. The computer-readable storage medium of claim 1 wherein the storage medium is a memory in a computing system.

15. The computer-readable storage medium of claim 1 wherein the data structure is accessible over a wide-area or local-area network.

16. A computer-implemented method for using a sentiment thesaurus stored according to the computer-readable storage medium of claim 1 to retrieve sentiment words for use with online content, comprising:

receiving an indication of an adjective;

determining from the sentiment thesaurus one or more related sentiment adjective word-senses that relate to the indicated adjective, by looking up the indicated adjective and retrieving related sentiment adjective word-senses; and

returning indications to the retrieved related sentiment adjective word-senses.

17. The method of claim 16 wherein the retrieved related sentiment adjectives are returned in an order.

18. The method of claim 17 wherein the order is based upon a connection type stored for the indicated adjective in relation to each related sentiment adjective word-sense.

19. The method of claim 17 wherein the order returns related sentiment adjective word-senses that are more closely related before those that are more distantly related, as determined based upon a set of connection rules.

20. The method of claim 16 wherein the looking up the indicated adjective and retrieving related sentiment adjective word-senses further comprises:

determining a set of related sentiment adjective word-senses that correspond to the indicated adjective based upon an indicated connection type or a range of connection types; and

returning the determined related sentiment adjective word-senses.

21. The method of claim 20 wherein the indicated connection type or the range of connection types indicates only connections where more specificity of meaning is occurring.

22. The method of claim 20 wherein the indicated connection type or the range of connection types indicates relationships in which the indicated adjective matches a valid sentiment adjective word sense or shares a sense with the valid sentiment adjective word sense.

23. The method of claim 16 wherein the returned indicated related sentiment adjective word-senses are presented in a list of viable sentiment words.

24. The method of claim 16 wherein one or more of the returned indicated related sentiment adjective word-senses are incorporated to tag the online content.

25. The method of claim 16 wherein the returned indicated related sentiment adjective word-senses are used for searching online content.

26. The method of claim 16 wherein the returned indicated related sentiment adjective word-senses are used to access and display content with opinion data reflecting one or more of the indicated related sentiment adjective word senses.

27. A computer-implemented method for facilitating the production of an online thesaurus used to generate sentiment data for online content, comprising:

generating and storing in a computer memory a plurality of valid sentiment adjective word-senses in a vocabulary data repository by determining whether each adjective from a set of possible adjectives is a valid sentiment based at least upon evaluation of a set of rules that define valid sentiments; and

for each generated valid sentiment adjective word-sense,

determining a set of related adjectives that meets a plurality of connection rules that limit, for each related adjective in the set, a degree of distance in meaning between the related adjective and the valid sentiment adjective word-sense; and

storing, in a thesaurus data structure in a computer memory, the set of related adjectives, along with indications of the plurality of connection rules that have been met, to correspond with the valid sentiment adjective word-sense, such that, responsive to an indication of one of the related adjectives, the valid sentiment adjective word-sense can be retrieved.

28. The method of claim 27 wherein the thesaurus data structure is stored in a table.

29. The method of claim 27 wherein the plurality of connection rules specify that two adjectives can be related if they represent no more than one shift in meaning.

30. The method of claim 27 wherein the plurality of connection rules specify that a related adjective can relate to a valid sentiment adjective word-sense as matching, as sharing a valid sense, as sharing a spoke-axle connection to a valid sense, as sharing an axle-axle connection to a valid sense, as sharing a spoke-axle-spoke connection to a valid sense, as sharing an axle-axle-spoke connection to a valid sense, as sharing a spoke-axle-axle connection to a valid sense, or as sharing a spoke-axle-axle-spoke connection to a valid sense.

31. The method of claim 27 wherein the set of rules used for determining whether each adjective from the set of possible adjectives is a valid sentiment comprises:

one or more of;

a rule that excludes a sense of the adjective if the adjective word sense does not have a meaning that can describe an individual piece of content;

a rule that excludes a sense of the adjective if the adjective word sense cannot describe a single noun in the online content without context provided by additional words of the online content;

a rule that excludes a sense of the adjective if the adjective word sense is a number;

a rule that excludes a sense of the adjective if the adjective word sense describes content from a third person perspective;

a rule that excludes a sense of the adjective if the adjective word sense is limited to the predicative or adjunctive state in common usage of the adjective;

a rule that excludes a sense of the adjective if the adjective word sense describes content non-subjectively in a manner that could describe all content;

a rule that excludes a sense of the adjective if the adjective word sense describes content as if existing as a tangible thing;

a rule that excludes a sense of the adjective if the adjective word sense describes content as related to a world-at-large;

a rule that excludes a sense of the adjective if the adjective word sense describes content in a sense of categorization;

a rule that excludes a sense of the adjective if the adjective word sense describes content in regard to formal structure of a content as a medium;

a rule that excludes a sense of the adjective if the adjective word sense is not a single word and is a lexical unit comprised of multiple aggregated words;

a rule that excludes a sense of the adjective if the adjective word sense describes content as undergoing or having undergone change;

a rule that excludes a sense of the adjective if the adjective word sense is an uncommon usage spelling variation of another adjective;

a rule that excludes a sense of the adjective if the adjective word sense has a negative prefix or suffix and has a meaning that is synonymous with an adjective already included in the vocabulary; and

/or

a rule that excludes a sense of the adjective if the adjective word sense anthropomorphizes corresponding content as a proxy for the author in a manner that fails to describe tonality of the corresponding content.