Method and apparatus for generating and/or updating cooccurrence relation dictionary

• US 5,181,163 A
• Filed: 07/31/1989
• Issued: 01/19/1993
• Est. Priority Date: 08/03/1988
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Nakajima, Hiroyuki, Kaji, Hiroyuki
• Primary Examiner(s): Smith, Jerry
• Assistant Examiner(s): Trammell, Jim
• Application Number:
  US07/387,598
• Time in Patent Office: 1,268 Days
• Field of Search: 364/419, 364/226.4, 364/975, 364/920.4
• US Class Current: 704/10
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G06F 40/247 Thesauruses; Synonyms
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: G06F 40/55 Rule-based translation

Abstract

A method of and an apparatus for generating and/or updating a cooccurrence relation dictionary in a natural language processing system such as a machine translation system including an input/output device, a memory for storing therein dictionary information, and a processor for processing an input sentence by use of the dictionary information. In the system, a syntactic analysis is performed on the input sentence, ambiguity is examined on each possible pair of words included in the result of the syntactic analysis, and only pairs of words determined to be free from ambiguity based on the result of the examination is converted into information of a predetermined format so as to register the converted information to the cooccurrence relation dictionary for a source language.

Claims

1. A method of generating and updating a cooccurrence relation directory in a natural language processing system including an input/output device, a memory for storing therein dictionary information, and a processor for processing an input sentence by use of the dictionary information, the method comprising the steps of:

effecting a syntactic analysis on the input sentence, wherein said syntactic analysis is effected such that all possible syntactic structures are produced, each structure being expressed as a set of predicate-argument relationships;

inspecting ambiguity in predicate-argument relationships included in said input sentence, wherein said ambiguity inspection is effected such that a predicate-argument relationship which occurs in said all possible syntactic structures produced by the syntactic analysis is judged to be unambiguous;

transforming unambiguous predicate-argument relationships obtained as a result of said inspecting into information having a predetermined format; and

registering said transformed unambiguous predicate-argument relationships to said cooccurrence relation dictionary.

2. A method of generating and updating a cooccurrence relation directory in a natural language processing system including an input/output device, a memory for storing therein dictionary information, and a processor for processing an input sentence by use of the dictionary information, the method comprising the steps of:

effecting a syntactic analysis on the input sentence;

inspecting ambiguity in predicate-argument relationships included in said input sentence;

transforming unambiguous predicate-argument relationships obtained as a result of said inspecting into information having a predetermined format;

registering said transformed unambiguous predicate-argument relationships to said cooccurrence relation dictionary; and

updating a case frame dictionary based on a set of generalizable predicate-argument relationships contained in the cooccurrence relation dictionary stored in said memory, said case frame dictionary including information of case relations satisfiable between a predicate word and an argument thereof.

3. A natural language processing system including an input/output device, file means for storing therein dictionary information, and processing means for processing an input sentence supplied from the input/output device, using the dictionary information,

said processing means comprising:

syntactic analysis means coupled to said input/output device for effecting a syntactic analysis on an input sentence to produce possible predicate-argument relationships;

ambiguity determining means responsive to said syntactic analysis means for examining ambiguity in said predicate-argument relationships;

a cooccurrence relation update means for registering in a cooccurrence relation dictionary storing cooccurrence relation information, predicate-argument relationships determined to be unambiguous by said ambiguity determining means, when said predicate-argument relationships are not included in said cooccurrence dictionary; and

case frame file means for constituting therein a case frame dictionary comprising information of a case relation satisfiable between a predicate word and a noun to be governed by the word;

said processing means referencing said cooccurrence relation dictionary file means and the case frame dictionary file means so as to automatically generalize knowledge accumulated in said cooccurrence relation dictionary file into case frame information, thereby updating said case frame dictionary.

4. The natural language processing system according to claim 3 wherein said system, further comprises:

translate means for generating from a first language of the input sentence, by use of a syntactic analysis result from said syntactic analysis means, an output sentence in a second language through a translation of the input sentence; and

mode select means for selectively executing each mode for said cooccurrence relation update means, said case frame dictionary update means, and said translate means.

5. A method of generating and updating a cooccurrence relation directory in a natural language processing system including an input/output device, a memory for storing therein dictionary information, and a processor for processing an input sentence by use of the dictionary information, the method comprising the steps of:

effecting a syntactic analysis on the input sentence;

inspecting syntactic ambiguity of said input sentence;

transforming syntactically unambiguous predicate-argument relationships obtained as a result of said inspecting into information having a predetermined format; and

registering said transformed unambiguous predicate-argument relationships to said cooccurrence relation dictionary.

6. A method according to claim 5, further comprising a step of generating an output sentence as a translation of the input sentence based on the result of the syntactic analysis.

7. A method according to claim 6, further including a step or receiving in advance from an external device select information specifying whether or not the output sentence generate step is to be executed,

said output sentence generate step being effected only when the select information specifies an execution thereof.

8. A method of automatically generating cooccurrence relation information indicating whether a pair of words in a natural language can cooccur using, a memory, sentences, and a processor,

said method comprising the steps of:

providing as an input, sentences belonging to a user selected field;

effecting a syntactic analysis of an input sentence having a set of possible predicate-argument relationships, on the basis of grammatical rules to determine an intermediate representation of the input sentence;

examining ambiguity in the predicate-argument relationships on the basis of predetermined rules to determine whether each predicate-argument relationship is ambiguous or not; and

selecting predicate-argument relationships which are determined to be syntactically unambiguous from the examination result by said examining, and providing the same as cooccurrence relation information to be registered in a cooccurrence relation dictionary storing cooccurrence relation information.

9. A method according to claim 8, further comprising updating said cooccurrence relation dictionary storing cooccurrence relation information in response to each providing of a pair of words determined not to have ambiguity.

10. A natural language processing system including an input/output device, file means for storing therein dictionary information, and processing means for processing an input sentence supplied from the input/output device, using the dictionary information.

said processing means comprising:

syntactic analysis means coupled to said input/output device for effecting syntactic analysis on an input sentence to produce possible syntactic structures, each of which is expressed as a set of predicate-argument relationships;

ambiguity determining means responsive to said syntactic analysis means for examining syntactic ambiguity in said predicate-argument relationships; and

a cooccurrence relation update means for registering in a cooccurrence relation dictionary storing cooccurrence relation information, predicate-argument relationships determined to be syntactically unambiguous by said ambiguity determining means, when said predicate-argument relationships are not included in said cooccurrence dictionary.

11. A natural language processing system according to claim 10, wherein said ambiguity determining means includes means for examining whether each pair of words is included in a set of all possible paris of words attained as an analysis result from said syntactic analysis means to determine only pairs of words commonly included therein to be free from ambiguity.

12. A method of supplying information to a cooccurrence relation dictionary of a natural language processing system, wherein the system includes an input/output device which inputs words forming a sentence, a memory for storing information supplied to the cooccurrence relation dictionary, and a processor for processing the input words, the method comprising the steps of:

inspecting the words input by the input device which are arranged as pairs of words in a predicate-argument relationship to determine whether the pairs of words are syntactically unambiguous;

registering automatically the pairs of words determined to be syntactically unambiguous to the cooccurrence relation dictionary; and

effecting a syntactic analysis on the input sentence.