CLASSIFICATION ENGINE FOR LEARNING PROPERTIES OF WORDS AND MULTI-WORD EXPRESSIONS

• US 20200134020A1
• Filed: 10/30/2018
• Published: 04/30/2020
• Est. Priority Date: 10/30/2018
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Rabinovich, Ella, Sznajder, Benjamin, Spector, Artem, Ilya, Shnayderman, AHARONOV, RANIT, KONOPNICKI, DAVID, Slonim, Noam
• Granted Patent: US 11,100,287 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/337 Profile generation, learning or modification
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

Method and apparatus for training and using a classifier for words. Embodiments include receiving a first plurality of sentences comprising a first word that is associated with a class and a second plurality of sentences comprising a second word that is not associated with the class. Embodiments include training a classifier using positive training data for the class that is based on the first plurality of sentences and negative training data for the class that is based on the second plurality of sentences. Embodiments include determining a measure of correlation between a third word and the class by using a sentence comprising the third word as an input to the classifier. Embodiments include using the measure of correlation to perform an action selected from the following list: selecting content to provide to a user; determining an automatic chat response; or filtering a set of content.

Claims

1-20. -20. (canceled)

21. A method comprising:

receiving a first plurality of sentences comprising a first word that is associated with a class and a second plurality of sentences comprising a second word that is not associated with the class;

training a classifier using positive training data for the class that is based on the first plurality of sentences and negative training data for the class that is based on the second plurality of sentences;

determining a measure of correlation between a third word and the class by using a sentence comprising the third word as an input to the classifier; and

using the measure of correlation to perform an action selected from the following list;

selecting content to provide to a user;

determining an automatic chat response;

or

filtering a set of content.

22. The method of claim 21, wherein training the classifier comprises identifying a context of the first word in the first plurality of sentences and a context of the second word in the second plurality of sentences.

23. The method of claim 22, wherein identifying the context of the first word in the first plurality of sentences and the context of the second word in the second plurality of sentences comprises identifying words near the first word in the first plurality of sentences and identifying words near the second word in the second plurality of sentences.

24. The method of claim 21, wherein the positive training data comprises one or more of the following:

a word;

a phrase;

a prefix;

or a suffix.

25. The method of claim 21, wherein the classifier comprises one of:

a nearest neighbor algorithm;

or a Naï

ve Bayes algorithm.

26. The method of claim 21, wherein selecting the content to provide to the user comprises identifying anecdotal content based on the measure of correlation.

27. The method of claim 21, wherein determining the automatic chat response comprises:

determining a difficulty of a problem based on the measure of correlation; and

choosing whether to escalate the problem to a live support agent based on the difficulty of the problem.

28. The method of claim 21, wherein the class comprises words defined as being abstract.

29. The method of claim 21, wherein training the classifier comprises filtering out non-alphabetic words and special characters from the first plurality of sentences and the second plurality of sentences.

30. The method of claim 21, wherein training the classifier comprises masking the first word in the first plurality of sentences and masking the second word in the second plurality of sentences.

31. A system, comprising:

one or more processors; and

a non-transitory computer-readable medium storing instructions that, when executed by the one or more processors, cause the one or more processors to perform a method, the method comprising;

receiving a first plurality of sentences comprising a first word that is associated with a class and a second plurality of sentences comprising a second word that is not associated with the class;

training a classifier using positive training data for the class that is based on the first plurality of sentences and negative training data for the class that is based on the second plurality of sentences;

determining a measure of correlation between a third word and the class by using a sentence comprising the third word as an input to the classifier; and

using the measure of correlation to perform an action selected from the following list;

selecting content to provide to a user;

determining an automatic chat response;

or

filtering a set of content.

32. The system of claim 31, wherein training the classifier comprises identifying a context of the first word in the first plurality of sentences and a context of the second word in the second plurality of sentences.

33. The system of claim 32, wherein identifying the context of the first word in the first plurality of sentences and the context of the second word in the second plurality of sentences comprises identifying words near the first word in the first plurality of sentences and identifying words near the second word in the second plurality of sentences.

34. The system of claim 31, wherein the positive training data comprises one or more of the following:

a word;

a phrase;

a prefix;

or a suffix.

35. The system of claim 31, wherein the classifier comprises one of:

a nearest neighbor algorithm;

or a Naï

ve Bayes algorithm.

36. The system of claim 31, wherein selecting the content to provide to the user comprises identifying anecdotal content based on the measure of correlation.

37. The system of claim 31, wherein determining the automatic chat response comprises:

determining a difficulty of a problem based on the measure of correlation; and

choosing whether to escalate the problem to a live support agent based on the difficulty of the problem.

38. The system of claim 31, wherein the class comprises words defined as being abstract.

39. The system of claim 31, wherein training the classifier comprises filtering out non-alphabetic words and special characters from the first plurality of sentences and the second plurality of sentences.

40. A computer program product comprising:

a computer-readable storage medium having computer-readable program code embodied therewith, the computer-readable program code executable by one or more computer processors to perform a method, the method comprising;

receiving a first plurality of sentences comprising a first word that is associated with a class and a second plurality of sentences comprising a second word that is not associated with the class;

training a classifier using positive training data for the class that is based on the first plurality of sentences and negative training data for the class that is based on the second plurality of sentences;

determining a measure of correlation between a third word and the class by using a sentence comprising the third word as an input to the classifier; and

using the measure of correlation to perform an action selected from the following list;

selecting content to provide to a user;

determining an automatic chat response;

or

filtering a set of content.