Ontology expansion using entity-association rules and abstract relations

Patent metadata

Field	Value
Publication number	US-11030406-B2
Application number	US-201615007703-A
Country	US
Kind code	B2
Filing date	Jan 27, 2016
Priority date	Jan 27, 2015
Publication date	Jun 8, 2021
Grant date	Jun 8, 2021

Abstract

A method for expanding an initial ontology via processing of communication data, wherein the initial ontology is a structural representation of language elements comprising a set of entities, a set of terms, a set of term-entity associations, a set of entity-association rules, a set of abstract relations, and a set of relation instances. A method for extracting a set of significant phrases and a set of significant phrase co-occurrences from an input set of documents further includes utilizing the terms to identify relations within the training set of communication data, wherein a relation is a pair of terms that appear in proximity to one another.

First claim

What is claimed is:

1. An automated method for developing and refining an initial ontology stored in computer-readable memory via processing of communication data, wherein the initial ontology is a structural representation of language elements, the method comprising:
developing the initial ontology from a first training set of communication data by a computing system, wherein the initial ontology comprises a set of entities, a set of terms, a set of term-entity associations, a set of entity-association rules, a set of abstract relations, and a set of relation instances;
generating a second training set of communication data by the computing system;
processing, by the computing system, the second training set of communication data to extract a set of significant phrases and a set of significant phrase pairs from within the second training set of communication data;
after the extracting, adding, by the computing system, the set of significant phrases as ontology terms to the set of terms stored in the initial ontology, and then associating, by the computing system, the added ontology terms to ontology entities of the set of entities;
after the associating, adding, by the computing system, new abstract relations to the set of abstract relations stored in the initial ontology;
after the adding of the new abstract relations to the stored initial ontology, adding, by the computing system, new relation instances to the set of relation instances stored in the initial ontology;
after completion of the adding of the new relation instances, outputting and storing the initial ontology, by the computing system, as a refined ontological structure;
receiving a plurality of transcribed interactions by the computing system;
loading the refined ontological structure into a memory of the computing system;
tagging the plurality of transcribed interactions using the set of entities, the set of terms, the set of term-entity associations, the set of entity-association rules, the set of abstract relations, and the set of relation instances of the loaded refined ontological structure; and
generating analytics based on the tagged plurality of transcribed interactions.

2. The method of claim 1 , wherein the computing system performs the steps of adding the set of significant phrases and associating the added ontology terms to ontology entities of the set of entities by at least sorting the set of significant phrases according to length, and then for each ϕ∈Φ performing at least the following:
a) if there exists a term T∈ comprising a word sequence that is identical to phrase ϕ, then phrase ϕ is not processed further;
b) if there exists a pair of terms T 1 ,T 2 ∈ that are mutually contained in phrase ϕ, and if there exists a pair of entities E 1 ,E 2 ∈ such that T 1 E 1 and T 2 E 2 —then phrase ϕ is not processed further;
c) otherwise, iterating over all entity-association rules , and compute:

X
*

⁢

=

arg
⁢

⁢
min

〈

E
1

,

E
2

,
E

〉

∈

⁢

X
*

⁢

{

d
⁡

(

T
1

,

E
1

)

+

d
⁡

(

T
2

,

E
2

)

|

T
1

,

T
2

⋐

ϕΛT
1

⋂

T
2

=
∅

}

;

i. if {tilde over (E)}={tilde over (0)}, discard the given phrase ϕ;
ii. if {tilde over (E)}={tilde over (1)}, add a new term T* into that corresponds to phrase ϕ and add a term-entity associations T*,E 1 into ;
iii. if {tilde over (E)}={tilde over (2)}, add a new term T* into that corresponds to phrase ϕ and add a term-entity associations T*,E 2 into ; and
iv. if {tilde over (E)}∈ , add a new term T* into that corresponds to phrase ϕ and add the term-entity associations T*,{tilde over (E)} ;
wherein Φ denotes the set of significant phrases and ϕ denotes a given phrase of the set of significant phrases Φ comprising a sequence of words w 1 , . . . , w n ;
wherein denotes the set of terms and T denotes a give term of the set of terms , wherein a given term T′ that comprises the word sequence w′ 1 , . . . , w′ k is contained in ϕ and is denoted by T′⊂ϕ if k<n and if there exists some index i such that: w i =w′ 1 , . . . , w i+k−1 =w′ k , wherein a given pair of terms T′,T″ are mutually contained in ϕ if both are contained in ϕ(T′,T″⊂ϕ) with no overlap between them (T′∩T″=Ø);
wherein denotes the set of entities and E denotes a given entity of the set of entities ;
wherein denotes the set of term-entity associations, wherein ⊆ × , wherein ∀T ∃E T,E ∈ ;
wherein a given distance between a term T and an entity E is denoted d(T,E), wherein d(T,E) is 0 i

Assignees

• Verint Systems Ltd

Inventors

• Baum Daniel
• Segal Uri
• Wein Ron
• Sidi Oana

Classifications

• G06F40/289 Primary

  Phrasal analysis, e.g. finite state techniques or chunking · CPC title

• G06F16/367 Primary

  Ontology · CPC title

• G06F40/284

  Lexical analysis, e.g. tokenisation or collocates · CPC title