EP0467527A2
Expired / Lapsed
EP
⏱️ 0.0 years remaining
Natural language apparatus and method and construction of a knowledge base for natural language analysis
Assignee: International Business Machines Corp
•
Filing: 1991-06-14
•
Granted: —
•
Expires: 2011-06-14
Abstract
Abstract A natural language analysis apparatus comprises: knowledge base means for storing first-type trees representing dependencies among words in sentences, and second-type trees representing taxonym relationships of words; table means responsive to entry of a word to output ID data of said first-type tree in which said word appears, node location data of said word in said first-type tree, and to output ID data of said second-type tree in which said word is contained as a hyponym; means for judging the structural ambiguity of an incoming sentence; means for extracting a candidate pair of modifier and modifiee for each possible dependency for a sentence judged to be ambiguous structurally; means for entering words comprising each said pair into said table means and for determining, on the basis of the output data, a path including said words at opposite ends and including some of the words appearing in the first-type tree; means for calculating a path distance for each said pair; and means for determining a most preferable dependency on the basis of said path distance calculated for each. <IMAGE>
Inventors
- 👤 Katashi Nagao
- 👤 Hiroshi Nomiyama
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-11-04 08:23
CPC Classifications
G; G06; G06F; G06F40/00; G06F40/20; G06F40/205; G06F40/211; G06F40/30
Claims Analysis
Total Claims: 13
Claim 1
Independent Claim
179 words
Apparatus for natural language analysis, the apparatus comprising: knowledge base means for storing first-type trees representing dependencies among words in sentences, and second-type trees representing taxonym relationships of words; table means responsive to entry of a word to output ID data of said first-type tree in which said word appears, node location data of said word in said first-type tree, and to output ID data of said second-type tree in which said word is contained as a hyponym; means for judging the structural ambiguity of an incoming sentence; means for extracting a candidate pair of modifier and modifiee for each possible dependency for a sentence judged to be ambiguous structurally; means for entering words comprising each said pair into said table means and for determining, on the basis of the output data, a path including said words at opposite ends and including some of the words appearing in the first-type tree; means for calculating a path distance for each said pair; and means for determining a most preferable dependency on the basis of said path distance calculated for each.
Claim 2
Dependent Claim
41 words
Apparatus as claimed in claim 1, further comprising: means for storing in said knowledge base means a first-type tree for the incoming sentence including said most preferable dependency determined by said means for determining and for renewing said table means responsively.
Claim 3
Dependent Claim
24 words
Apparatus as claimed in claim 2 wherein said knowledge base means separately stores learned data and context data added by said means for storing.
Claim 4
Dependent Claim
20 words
Apparatus as claimed in claim 2 wherein said table means is separately prepared for learned data and for context data.
Claim 5
Dependent Claim
24 words
Apparatus as claimed in claim 1 wherein said means for calculating calculates said distance, based on the number of dependencies included in the path.
Claim 6
Dependent Claim
19 words
Apparatus as claimed in claim 1 wherein said first-type tree is provided with semantic case data for each dependency.
Claim 7
Dependent Claim
36 words
Apparatus as claimed in claim 6 wherein said means for calculating calculates said distance according to the consistency between the case relationship between a modifier and a candidate modifiee and the case relationship for the path.
Claim 8
Dependent Claim
41 words
Apparatus as claimed in claim 1 wherein said means for calculating calculates said distance, on the basis of the consistency of co-occurrence of a word included in said incoming sentence and a word included in said first-type tree for the path.
Claim 9
Dependent Claim
35 words
Apparatus as claimed in claim 1 wherein said means for calculating calculates said distance, on the basis of the degree of consistency between the path and a first-type tree. added by said means for determining.
Claim 10
Dependent Claim
64 words
Apparatus as claimed in claim 1 wherein said second-type tree is an "isa" tree having only two nodes corresponding to a hypernym and a hyponym, and wherein said means for entering is responsive to an output of a hypernym of a word forming the pair, to iterate search for an "isa" tree including said hypernym as a hyponym, thereby producing a chain of hypernyms.
Claim 11
Dependent Claim
17 words
Apparatus as claimed in claim 1 wherein wherein a synonym relationship is represented by two "isa" trees.
Claim 12
Dependent Claim
181 words
In a computer system including a knowledge base that stores first-type trees representing dependencies among words in sentences and second-type trees representing taxonym relationships of words, and including a table responsive to entry of a word for outputting ID data of said first-type tree in which said word appears, node location data of said word in said first-type tree, and ID data of said second-type tree in which said word appears as a hyponym, a natural language analysis method comprising the steps of: (a) judging the structural ambiguity of an incoming sentence; (b) extracting a candidate pair of modifier and modifiee for each possible dependency as for a sentence judged to be structurally ambiguous; (c) entering words comprising each pair into said table means and determining, on the basis of the output data, a path that has said words at opposite ends and contains some of the words appearing in said first-type tree; (d) calculating a path distance for each pair; and (e) determining the most preferable dependency relationship, on the basis of said path distance calculated for each said pair.
Claim 13
Independent Claim
104 words
A method of constructing a knowledge base for natural language analysis comprising the steps of: (a) preparing a knowledge base for storing trees representing dependencies among words in sentences and preparing a table responsive to entry of a word for outputting ID data of a tree containing said word and node location data of said word in said tree; (b) determining the most preferable of the possible dependencies for an incoming sentence by using said knowledge base and said table; and (c) storing in said knowledge base a tree for the incoming sentence that includes said most preferable dependency and renewing said table responsively.
References Cited (Backward Citations)
No citation records indexed.
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.