Extraction and grouping of feature words

• US 8,484,228 B2
• Filed: 03/17/2011
• Issued: 07/09/2013
• Est. Priority Date: 03/17/2011
• Status: Active Grant

• Current Assignee: Indian Institute of Science, SAP SE
• Original Assignee: Indian Institute of Science, SAP AG (SAP SE)
• Inventors: Bhattacharyya, Chiranjib, Lakkaraju, Himabindu, Nath, Kaushik, Arvindam, Sunil
• Primary Examiner(s): HO, BINH VAN
• Application Number:
  US13/049,922
• Publication Number:
  US 20120239668A1
• Time in Patent Office: 845 Days
• Field of Search: 707/754
• US Class Current: 707/754
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06Q 10/10 Office automation; Time management

Abstract

Various embodiments of systems and methods for extraction and grouping of feature words are described herein. Feature words are obtained from a first corpus of text bodies comprising a plurality of reviews. A second corpus is created using a combination of the obtained feature words, verbs and adjectives from the first corpus. The second corpus comprises filtered reviews and each of the filtered reviews pertains to a review. Topics are preliminarily assigned for words in the filtered reviews of the second corpus. For each of the feature words in the second corpus, a topic count is determined for every preliminarily assigned topic. After determining the topic count, one or more of the topics are finally assigned to the feature words based on a topic count value. At least one topic is presented as a group of the feature words for which the at least one topic is assigned based on the topic count value.

Claims

1. An article of manufacture including a non-transitory computer readable storage medium to tangibly store instructions, which when executed by a computer, cause the computer to:

obtain feature words from a first corpus of text bodies comprising a plurality of reviews;

create a second corpus using a combination of the obtained feature words, verbs and adjectives from the first corpus, wherein the second corpus comprises filtered reviews and each of the filtered reviews pertains to a review of the plurality of reviews;

preliminarily assign topics for words in the filtered reviews of the second corpus;

for the feature words in the second corpus, determine topic counts for preliminarily assigned topics;

after determining the topic counts, finally assign one or more of the topics to the feature words based on a topic count value, comprising;

assign the one or more topics that have a highest topic count value to the featuring words; and

assign the one or more topics that have a threshold topic count value relative to the highest topic count value; and

present on a user interface at least one topic as a group of the feature words for which the at least one topic is assigned based on the topic count value.

2. The article of manufacture of claim 1, wherein the instructions to obtain the feature words from the first corpus of text bodies, further comprises instructions to:

extract feature words from the first corpus;

obtain opinion fragments from the first corpus;

determine a relative occurrence for each of the feature words based on occurrence of the feature words within the opinion fragments and within the first corpus; and

retain one or more feature words having corresponding relative occurrences more than a predefined threshold.

3. The article of manufacture of claim 2, further comprising instructions, which when executed by a computer, cause the computer to:

consider the feature words as occurring within the opinion fragments even when the feature words are referred in the opinion fragments by anaphora resolution.

4. The article of manufacture of claim 1, wherein the instructions to determine the topic count, further comprises instructions to:

determine a particular topic count as an aggregate of the number of assignments of a particular topic to a particular feature word across all the filtered reviews in the second corpus.

5. The article of manufacture of claim 1, wherein the first corpus is a non-generic corpus of text bodies.

6. The article of manufacture of claim 1, wherein the feature words comprise compound words.

7. A computer-implemented method for grouping words, the method comprising:

obtaining feature words from a first corpus of text bodies comprising a plurality of reviews;

creating a second corpus using a combination of the obtained feature words, verbs and adjectives from the first corpus, wherein the second corpus comprises filtered reviews and each of the filtered reviews pertains to a review of the plurality of reviews;

preliminarily assigning topics for words in the filtered reviews of the second corpus;

for the feature words in the second corpus, determining topic counts for preliminarily assigned topics;

after determining the topic counts, finally assigning one or more of the topics to the feature words based on a topic count value, comprising;

assigning the one or more topics that have a highest topic count value to the feature words; and

assigning the one or more topics that have a threshold topic count value relative to the highest topic count value; and

presenting on a user interface at least one topic as a group of the feature words for which the at least one topic is assigned based on the topic count value.

8. The method of claim 7, wherein obtaining the feature words from the first corpus of text bodies, further comprises:

extracting feature words from the first corpus;

obtaining opinion fragments from the first corpus;

determining a relative occurrence for each of the feature words based on occurrence of the feature words within the opinion fragments and within the first corpus; and

retaining one or more feature words having corresponding relative occurrences more than a predefined threshold.

9. The method of claim 8, further comprises:

considering the feature words as occurring within the opinion fragments even when the feature words are referred in the opinion fragments by anaphora resolution.

10. The method of claim 7, wherein determining the topic count comprises:

determining a particular topic count as an aggregate of the number of assignments of a particular topic to a particular feature word across all the filtered reviews in the second corpus.

11. The method of claim 7, wherein the first corpus is a non-generic corpus of text bodies.

12. The method of claim 7, wherein the feature words comprise compound words.

13. A computer system for grouping words, comprising:

a computer memory to store program code; and

a processor to execute the program code to;

obtain feature words from a first corpus of text bodies comprising a plurality of reviews;

create a second corpus using a combination of the obtained feature words, verbs and adjectives from the first corpus, wherein the second corpus comprises filtered reviews and each of the filtered reviews pertains to a review of the plurality of reviews;

preliminarily assign topics for words in the filtered reviews of the second corpus;

for the feature words in the second corpus, determine topic counts for preliminarily assigned topics;

after determining the topic counts, finally assign one or more of the topics to the feature words based on a topic count value, comprising;

assign the one or more topics that have a highest topic count value to the feature words; and

assign the one or more topics that have a threshold topic count value relative to the highest topic value; and

present on a user interface at least one topic as a group of the feature words for which the at least one topic is assigned based on the topic count value.

14. The system of claim 13, wherein the program code to obtain the feature words from the first corpus of text bodies, further comprises program code to:

extract feature words from the first corpus;

obtain opinion fragments from the first corpus;

determine a relative occurrence for each of the feature words based on occurrence of the feature words within the opinion fragments and within the first corpus; and

retain one or more feature words having corresponding relative occurrences more than a predefined threshold.

15. The system of claim 14, wherein the processor further executes the program code to:

consider the feature words as occurring within the opinion fragments even when the feature words are referred in the opinion fragments by anaphora resolution.

16. The system of claim 13, wherein the program code to determine the topic count, further comprises program code to:

determine a particular topic count as an aggregate of the number of assignments of a particular topic to a particular feature word across all the filtered reviews in the second corpus.

17. The system of claim 13, wherein the first corpus is a non-generic corpus of text bodies.

18. The system of claim 13, wherein the feature words comprise compound words.