SYSTEM AND METHOD DETECTING HIDDEN CONNECTIONS AMONG PHRASES

• US 20160026617A1
• Filed: 10/01/2015
• Published: 01/28/2016
• Est. Priority Date: 03/24/2010
• Status: Active Grant

• Current Assignee: Innovid
• Original Assignee: Taykey Ltd. (Innovid Corp.)
• Inventors: Avner, Amit, Dror, Omer, BIRNBOIM, Itay
• Granted Patent: US 10,268,670 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06F 40/205 Parsing

Abstract

A system and method for identifying hidden connections among non-sentiment phrases are presented. The method includes identifying all connections among a plurality of non-sentiment phrases based on at least one proximity rule; determining direct connections among the identified connections, wherein each direct connection meets a predetermined correlation; filtering out the determined direct connections from the identified connections to yield hidden connections among the identified connections; analyzing the hidden connections to identify a common phrase, wherein the common phrase is associated with at least two hidden connections; generating a new hidden connection among the plurality of non-sentiment phrases based on the common phrase; and associating a sentiment phrase with at least two non-sentiment phrases having a hidden connection, wherein the association is a term taxonomy.

Claims

1. A method for identifying hidden connections among non-sentiment phrases, comprising:

identifying all connections among a plurality of non-sentiment phrases based on at least one proximity rule;

determining direct connections among the identified connections, wherein each direct connection meets a predetermined correlation;

filtering out the determined direct connections from the identified connections to yield hidden connections among the identified connections;

analyzing the hidden connections to identify a common phrase, wherein the common phrase is associated with at least two hidden connections;

generating a new hidden connection among the plurality of non-sentiment phrases based on the common phrase; and

associating a sentiment phrase with at least two non-sentiment phrases having a hidden connection, wherein the association is a term taxonomy.

2. The method of claim 1, further comprising:

extracting a plurality of phrases from textual content; and

identifying a plurality of sentiment phrases and the plurality of non-sentiment phrases in the plurality of phrases, wherein the associated sentiment phrase is from the plurality of sentiment phrases.

3. The method of claim 2, wherein identifying a plurality of sentiment phrases and the plurality of non-sentiment phrases in the plurality of phrases further comprises:

comparing each extracted phrase to a plurality of predetermined sentiment phrases and non-sentiment phrases; and

for each extracted phrase, determining whether the extracted phrase is a sentiment phrase or a non-sentiment phrase, wherein the extracted phrase is determined to be a sentiment phrase if the extracted phrase matched one of the predetermined sentiment phrases, wherein further the extracted phrase is determined to be a non-sentiment phrase if the extracted phrase matched one of the predetermined non-sentiment phrases.

4. The method of claim 3, further comprising:

assigning a sentiment score to each determined sentiment phrase, wherein the sentiment score represents a degree of any of;

positivity, neutrality, and negativity.

5. The method of claim 4, further comprising:

continuously monitoring feedback related to the determined sentiment phrases to identify a change in degree; and

assigning a new sentiment score based on the change in degree.

6. The method of claim 3, wherein each sentiment score is weighted respective of each other sentiment score.

7. The method of claim 2, wherein identifying a plurality of sentiment phrases and the plurality of non-sentiment phrases in the plurality of phrases further comprises any of:

replacing slang phrases with corresponding standard language phrases, removing conjunctions from phrases, removing unknown words, removing repeated words, and removing irrelevant words.

8. The method of claim 1, wherein the proximity rule includes at least one of:

a number of mentions of at least two non-sentiment phrases in a web page, a number of mentions of at least two non-sentiment phrases in different web pages, a number of mentions of at least two non-sentiment phrases in a piece of collected textual content, and a number of web pages within a web site between at least two mentioned non-sentiment phrases.

9. The method of claim 1, wherein each hidden connection is any of:

a first degree separation between at least two non-sentiment phrases, and a first degree separation between at least two different non-sentiment phrases that are associated with a common non-sentiment phrase.

10. The method of claim 1, wherein at least two non-sentiment phrases have a connection when they meet the at least one proximity rule.

11. The method of claim 1, wherein each hidden connection is between at least two non-sentiment phrases.

12. The method of claim 1, wherein generating a new hidden connection further comprises:

removing the identified common phrase.

13. A non-transitory computer readable medium having stored thereon instructions for causing one or more processing units to execute the method according to claim 1.

14. A system for identifying hidden connections among non-sentiment phrases, comprising:

a processor; and

a memory, the memory containing instructions that, when executed by the processor, configure the system to;

identify all connections among a plurality of non-sentiment phrases based on at least one proximity rule;

determine direct connections among the identified connections, wherein each direct connection meets a predetermined correlation;

filter out the determined direct connections from the identified connections to yield hidden connections among the identified connections;

analyze the hidden connections to identify a common phrase, wherein the common phrase is associated with at least two hidden connections;

generate a new hidden connection among the plurality of non-sentiment phrases based on the common phrase; and

associate a sentiment phrase with at least two non-sentiment phrases having a hidden connection, wherein the association is a term taxonomy.

15. The system of claim 14, wherein the system is further configured to:

extract a plurality of phrases from textual content; and

identify a plurality of sentiment phrases and the plurality of non-sentiment phrases in the plurality of phrases, wherein the associated sentiment phrase is from the plurality of sentiment phrases.

16. The system of claim 15, wherein identifying a plurality of sentiment phrases and the plurality of non-sentiment phrases in the plurality of phrases further comprises:

compare each extracted phrase to a plurality of predetermined sentiment phrases and non-sentiment phrases; and

for each extracted phrase, determine whether the extracted phrase is a sentiment phrase or a non-sentiment phrase, wherein the extracted phrase is determined to be a sentiment phrase if the extracted phrase matched one of the predetermined sentiment phrases, wherein further the extracted phrase is determined to be a non-sentiment phrase if the extracted phrase matched one of the predetermined non-sentiment phrases.

17. The system of claim 16, wherein the system is further configured to:

assign a sentiment score to each determined sentiment phrase, wherein the sentiment score represents a degree of any of;

positivity, neutrality, and negativity.

18. The system of claim 15, wherein the system is further configured to:

continuously monitor feedback related to the determined sentiment phrases to identify a change in degree; and

assign a new sentiment score based on the change in degree.

19. The system of claim 16, wherein each sentiment score is weighted respective of each other sentiment score.

20. The system of claim 15, the system is further configured to perform any of:

replace slang phrases with corresponding standard language phrases, remove conjunctions from phrases, remove unknown words, remove repeated words, and remove irrelevant words.

21. The system of claim 14, wherein the proximity rule includes at least one of:

a number of mentions of at least two non-sentiment phrases in a web page, a number of mentions of at least two non-sentiment phrases in different web pages, a number of mentions of at least two non-sentiment phrases in a piece of collected textual content, and a number of web pages within a web site between at least two mentioned non-sentiment phrases.

22. The system of claim 14, wherein each hidden connection is any of:

a first degree separation between at least two non-sentiment phrases, and a first degree separation between at least two different non-sentiment phrases that are associated with a common non-sentiment phrase.

23. The system of claim 14, wherein at least two non-sentiment phrases have a connection when they meet the at least one proximity rule.

24. The system of claim 14, wherein each hidden connection is between at least two non-sentiment phrases.

25. The system of claim 14, wherein the system is further configured to:

remove the identified common phrase.