Sentiment Analyzer

• US 20130173254A1
• Filed: 12/22/2012
• Published: 07/04/2013
• Est. Priority Date: 12/31/2011
• Status: Abandoned Application

• Current Assignee: Farrokh Alemi
• Original Assignee: Farrokh Alemi
• Inventors: Alemi, Farrokh
• Application Number:
  US13/726,059
• Publication Number:
  US 20130173254A1
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking
• CPC Class Codes: G06F 40/40 Processing or translation of natural language natural language analysis G06F40/20; semantic analysis G06F40/30

Abstract

A sentiment analysis tool receives a t-gram from an electronic device. The t-gram comprises gram(s), each of the gram(s) representing a word in a collection of words. A polarity is set for the t-gram. Possible smaller-gram combinations are generated from the t-gram. Until a condition is met, iterative actions are taken. A likelihood ratio is calculated for the largest of the smaller-gram combinations employing the training set. A determination is made of whether the likelihood ratio meets a minimum replication threshold. If satisfied: the smaller-gram combinations most distant from an undefined polarity value are selected, the smaller-gram combinations employed in calculating the likelihood ratio are excluded; the polarity value for the t-gram is increasing proportional to the likelihood ratio; and the training set is reduced to v-grams that include the t-gram. Otherwise, the size of the smaller-gram is reduced by 1.

Claims

1. A non-transitory tangible computer readable media containing one or more instructions executable by one or more processors to perform the method comprising:

a. receiving a t-gram from an electronic device, the t-gram comprising one or more grams, each of the one or more grams representing a word in a collection of words;

b. setting a polarity for the t-gram;

c. generating possible smaller-gram combinations from the t-gram;

d. iteratively, while the largest of the smaller-gram combinations is larger than zero and the number of v-grams in a training set is above a threshold;

i. calculating a likelihood ratio for the largest of the smaller-gram combinations employing the training set; and

ii. determining if the likelihood ratio meets a minimum replication threshold;

1. if the minimum replication threshold is satisfied;

a. selecting the smaller-gram combinations that is most distant from an undefined polarity value;

b. excluding the smaller-gram combinations employed in calculating the likelihood ratio;

c. increasing the polarity value for the t-gram proportional to the likelihood ratio; and

d. reducing the training set to v-grams that include the t-gram; and

2. if the minimum replication threshold is not satisfied, reducing the size of the smaller-gram by 1; and

e. reporting, via the electronic device, the polarity value.

2. The media according to claim 1, wherein a negative polarity corresponds to at least one of the following:

a. a negative sentiment;

b. a complaint;

c. an absence of an event; and

d. absence of a matched word in a dictionary.

3. The media according to claim 1, wherein a positive polarity corresponds to at least one of the following:

a. a positive sentiment;

b. a compliment;

c. an presence of an event; and

d. a matched word in a dictionary.

4. The media according to claim 1, wherein the likelihood ratio is the prevalence of the t-gram in a positive polarity in a training set divided by the prevalence of the t-gram in a negative polarity in a training set.

5. The media according to claim 1, wherein the likelihood ratio of the t-gram is zero or infinity, the likelihood ratio is calculated based on at least one of the following:

a. the size of a training set; and

b. the number of times the t-gram is classified with a negative polarity.

6. The media according to claim 1, wherein the training set comprises a set of v-grams with known polarities.

7. The media according to claim 1, wherein the training set comprises a set of v-grams with different sizes m.

8. The media according to claim 1, wherein the undefined polarity value is one.

9. The media according to claim 1, wherein the generating possible smaller-gram combinations from the t-gram further includes generating all possible smaller-gram combinations from the t-gram.

10. The media according to claim 1, wherein the generating possible smaller-gram combinations from the t-gram further includes generating possible smaller-gram combinations from the t-gram in the order of size.

11. The media according to claim 1, wherein the generating possible smaller-gram combinations from the t-gram further includes generating possible smaller-gram combinations employing consecutive grams in the t-gram.

12. The media according to claim 1, wherein the generating possible smaller-gram combinations from the t-gram further includes generating possible smaller-gram combinations employing non-consecutive grams in the t-gram comprising consecutive grams with at least one skipped gram.

13. The media according to claim 1, further including setting a minimum replication threshold as the maximum of:

a. the expected number of t-grams that can be randomly observed based on the independent occurrence of each gram; and

b. the number of v-grams in the training set needed to detect a significant difference between the observed occurrence of the t-gram and an uncertainty value.

14. The media according to claim 1, wherein the determining if the t-gram meets a minimum replication threshold comprises calculating the expected number of t-grams that can be randomly observed based on the independent occurrence of each gram.

15. The media according to claim 1, wherein the determining if the t-gram meets a minimum replication threshold comprises calculating the number of v-grams in the training set needed to detect a significant difference between the observed occurrence of the t-gram and an uncertainty value.

16. The media according to claim 15, wherein the uncertainty value if 0.5.

17. The media according to claim 1, wherein the electronic device is one of the following:

a. a tablet;

b. a computer;

c. a cell phone;

d. a mobile computing device; and

e. a server.

18. The media according to claim 1, wherein the setting the polarity of the collection of t-grams includes setting a polarity variable to 1.

19. The media according to claim 1, further comprising performing real time performance evaluations through examining time to or events until next negative polarity.

20. The media according to claim 1, wherein the word is one of the following:

a. a representation of an element in a spoken language;

b. a representation of an element in a written language;

c. a representation of an element in a computer language;

d. a representation of biological element in a series of biological elements;

e. a nucleotide in a strand of DNA; and

f. an event in a series of consecutive events.

21. The media according to claim 1, further comprising converting an audio t-gram recorded using an electronic recording device comprising spoken words separated by pauses to a text t-gram.