Automated ontology development

• US 10,339,452 B2
• Filed: 02/05/2014
• Issued: 07/02/2019
• Est. Priority Date: 02/06/2013
• Status: Active Grant

• Current Assignee: Verint Systems Incorporated
• Original Assignee: Verint Systems Limited (Verint Systems Incorporated)
• Inventors: Romano, Roni, Horesh, Yair, Dreyfuss, Jeremie
• Primary Examiner(s): Opsasnick, Michael N
• Application Number:
  US14/173,435
• Publication Number:
  US 20140222419A1
• Time in Patent Office: 1,973 Days
• Field of Search: None
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition

Abstract

Systems and methods of automated ontology development include a corpus of communication data. The corpus of communication data includes communication data from a plurality of interactions and is processed. A plurality of terms are extracted from the corpus. Each term of the plurality is a plurality of words that identify a single concept within the corpus. An ontology is automatedly generated from the extracted terms.

Claims

1. A method of automated ontology development via a computer system for processing communication data, wherein the ontology is a structural representation of language elements and the relationships between those language elements within a domain stored in memory of the computer system, the method comprising:

processing a corpus of communication data by the computer system, the corpus comprising communication data from a plurality of interactions from multiple platforms;

extracting a plurality of terms from the corpus by the computer system, wherein each term of the plurality is a plurality of words that identify a single concept within the corpus;

automatedly generating an ontology by the computer system from the extracted terms by at least creating two context vectors for each of the plurality of terms and comparing the context vectors for each of the plurality of terms to one another to categorize the terms into a plurality of relations, wherein a first of the two context vectors of a given term is a first list of terms that predicts terms that will appear to the left of the given term, wherein a second of the two context vectors is a second list of terms that predicts terms that will appear to the right of the given term, wherein each of the context vectors includes up to a predetermined number of potential terms in the first or second list of terms taken in descending order based on a calculated score predictive of the likelihood that a specific term will appear adjacent to a given one of the plurality of terms within the meaning units; and

storing the automatedly generated ontology in an ontology database in the memory of the computer system.

2. The method of claim 1, wherein processing the corpus further comprises:

receiving raw communication data; and

applying a rank filter to select a portion of the raw communication data as the corpus of communication data.

3. The method of claim 2, wherein the raw communication data comprises transcriptions of interactions, agent scripts, service manuals, and product manuals.

4. The method of claim 1, wherein processing the corpus further comprises:

identifying scripts within the corpus, wherein scripts are recurring patterns of three or more words.

5. The method of claim 4, wherein processing the corpus further comprises:

Zoning the communication data to segment the communication data into meaning units.

6. The method of claim 5, wherein the plurality of terms are extracted from the corpus on a meaning unit-by-meaning unit basis.

7. The method of claim 1, wherein the plurality of interactions are customer service interactions and the ontology is tailored for use in analyzing customer service interactions.

8. The method of claim 1, wherein the ontology comprises the plurality of terms, the plurality of relations, and a plurality of themes identified from the corpus.

9. The method of claim 1, wherein automatedly generating the ontology further comprises:

comparing the plurality of relations to one another to categorize the relations into a plurality of themes.

17. The method of claim 2, wherein the raw communication data comprises data from the interactions from multiple platforms including interactions made via one or more of by phone, email, internet chat, text message, web page comment, social media interaction, customer surveys, an audio recording, streaming audio, a transcription of spoken content, or written correspondence.

18. The method of claim 17, wherein the rank filter selects data files from the raw communication data that include a threshold of identified related terms to the domain of the ontology that is to be developed.

10. A method of automated ontology development via a computer system for processing communication data, wherein the ontology is a structural representation of language elements and the relationships between those language elements within a domain stored in memory of the computer system, the method comprising:

processing a corpus of communication data by the computer system, the corpus comprising communication data from a plurality of interactions from multiple platforms, by zoning the communication data to segment the communication data into a plurality of meaning units;

extracting a plurality of terms from each of the plurality of meaning units by the computer system, wherein each term of the plurality is a plurality of words that identify a single concept within the corpus, wherein words in a given one of the meaning units are assigned to only one term;

automatedly generating an ontology by the computer system that comprises the extracted terms by at least creating two context vectors for each of the plurality of terms and comparing the context vectors for each of the plurality of terms to one another to categorize the terms into a plurality of relations, wherein a first of the two context vectors of a given term is a first list of terms that predicts terms that will appear to the left of the given term, wherein a second of the two context vectors is a second list of terms that predicts terms that will appear to the right of the given term, wherein each of the context vectors includes up to a predetermined number of potential terms in the first or second list of terms taken in descending order based on a calculated score predictive of the likelihood that a specific term will appear adjacent to a given one of the plurality of terms within the meaning units; and

storing the automatedly generated ontology in an ontology database.

11. The method of claim 10, wherein processing the corpus further comprises:

receiving raw communication data; and

applying a rank filter to select a portion of the raw communication data as the corpus of communication data.

12. The method of claim 10, wherein automatedly generating the ontology further comprises:

comparing the plurality of relations to one another to categorize the relations into a plurality of themes.

13. The method of claim 12, wherein the ontology further comprises the plurality of relations and the plurality of themes.

19. The method of claim 11, wherein the rank filter selects data files from the raw communication data that include a threshold of identified related terms to the domain of the ontology that is to be developed.

20. The method of claim 19, wherein the raw communication data comprises data from the interactions from multiple platforms including interactions made via one or more of by phone, email, internet chat, text message, web page comment, social media interaction, customer surveys, an audio recording, streaming audio, a transcription of spoken content, or written correspondence.

14. A system for automated ontology development, wherein the ontology is a structural representation of language elements and the relationships between those language elements within a domain stored in memory, the system comprising:

a communication data database populated with communication data from interactions from multiple platforms;

a processor communicatively connected to the database of communication data and communicatively connected to a computer readable medium programmed with computer readable code that upon execution by the processor causes the processor to;

process a corpus of communication data received from the database through application of a rank filter to select the communication data from raw communication data, wherein the rank filter selects data files from the raw communication data that include a threshold of identified related terms to the domain of the ontology that is to be developed;

zone the communication data to segment the communication data into a plurality of meaning units;

extract a plurality of terms from the corpus on a meaning unit-by-meaning unit basis, wherein each term of the plurality is a plurality of words that identify a single concept within the corpus, wherein words in a given one of the meaning units are assigned to only one term; and

automatedly generate the ontology from the extracted terms by at least creating two context vectors for each of the plurality of terms and comparing the context vectors for each of the plurality of terms to one another to categorize the terms into a plurality of relations, wherein a first of the two context vectors of a given term is a first list of terms that predicts terms that will appear to the left of the given term, wherein a second of the two context vectors predicts terms is a second list of terms that that will appear to the right of the given term, wherein each of the context vectors includes up to a predetermined number of potential terms in the first or second list of terms taken in descending order based on a calculated score predictive of the likelihood that a specific term will appear adjacent to a given one of the plurality of terms within the meaning units; and

an ontology database upon which the processor stores the automatedly generated ontology.

15. The system of claim 14, wherein the communication data comprises transcriptions of interactions, agent scripts, service manuals, and product manuals.

16. The system of claim 14, further comprising:

a script database communicatively connected to the processor; and

wherein execution of the computer readable code by the processor further causes the processor to;

surface a plurality of scripts from the communication data;

store the plurality of scripts at the script database; and

apply the plurality of scripts from the script database to the corpus of communication data to identify scripts within the corpus of communication data.