Type-based selection of rules for semantically disambiguating words

• US 6,405,162 B1
• Filed: 09/23/1999
• Issued: 06/11/2002
• Est. Priority Date: 09/23/1999
• Status: Expired due to Term

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Segond, Frédérique, Brun, Caroline
• Primary Examiner(s): EDOUARD, PATRICK NESTOR
• Application Number:
  US09/401,685
• Time in Patent Office: 992 Days
• Field of Search: 704/1, 704/9, 704/10, 707/530, 707/531, 707/532, 707/5, 707/6
• US Class Current: 704/9
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

In semantically disambiguating words, where more than one disambiguation applies to the context in which a word occurs, a rule can be selected based on the type of information from which it was obtained. The rules can be derived from different types of information in a corpus such as a dictionary, and rules can be selected in accordance with a prioritization of the types of information.

Claims

1. A method of semantically disambiguating words using rules, the rules including rules derived from two or more types of information in a corpus, rules derived from at least two of the types being applicable to words occurring in specified contexts;

the method comprising;

(A) obtaining context information about a context in which a semantically ambiguous word occurs in an input text;

a first rule derived from a first type of corpus information and a second rule derived from a second type of corpus information both being applicable to words occurring in the context; and

(B) selecting the first rule rather than the second rule to disambiguate the semantically ambiguous word according to a selection order based on the types of corpus information from which the rules are derived.

2. The method of claim 1 in which each of the rules in the first and second types includes a context descriptor specifying contexts in which the rule is applicable.

3. The method of claim 2 in which each context descriptor includes two or more word descriptors;

each word descriptor in the first and second rules specifying one normalized word form.

4. The method of claim 2 in which each context descriptor includes two or more word descriptors;

the word descriptors in each of the first and second rules including a word descriptor that specifies an ambiguity class of word categories.

5. A method of semantically disambiguating words using rules, the rules including rules derived from two or more types of information in a corpus, rules derived from at least two of the types being applicable to words occurring in specified contexts, each of the rules in first and second types includes a context descriptor specifying contexts in which the rule is applicable, each context descriptor includes two or more word descriptors and a relation descriptor specifying a type of relation in which words that satisfy the word descriptors can occur;

the method comprising;

(A) using the input text to obtain relation information about a set of relations between the semantically ambiguous word and other words in the input text and, for each relation, word information about words that occur in the relation in the input text;

the context information including the relation information and the word information; and

(B) for each of the rules comparing the rule'"'"'s context descriptor with the context information and obtaining match information indicating that the context descriptors of the first and second rules are both satisfied by a relation between the semantically ambiguous word and other words in the input text;

in response to the match information, comparing the types of corpus information from which the first and second rules were derived and determining that the first type of corpus information has higher priority than the second type of corpus information; and

selecting to disambiguate the semantically ambiguous word using the first rule rather than the second rule based on the determination that the first type of corpus information has higher priority.

6. The method of claim 5 in which the corpus is a dictionary, the dictionary including a set of types of information that include the first and second types of corpus information;

the types in the set including, from highest to lowest priority, collocates, idioms, compounds, structure examples, phrasal verb examples, usages, and general examples.

7. A method of semantically disambiguating words using rules, the rules including rules derived from two or more types of information in a corpus, rules derived from at least two of the types being applicable to words occurring in specified contexts, each of the rules in first and second types includes a context descriptor specifying contexts in which the rule is applicable, each context descriptor includes two or more word descriptors and a relation descriptor specifying a type of relation in which words that satisfy the word descriptors can occur;

the method comprising;

(A) using the input text to obtain relation information about a set of relations between the semantically ambiguous word and other words in the input text and, for each relation, word information about words that occur in the relation in the input text;

the context information including relation information and the word information; and

(B) for each of the rules, comparing the rule'"'"'s context descriptor with the context information and obtaining match information indicating that the context descriptors of the first and second rules and a third rule are all satisfied by a relation between the semantically ambiguous word and other words in the input text;

the third rule also being derived from the first type of corpus information; and

in response to the match information, comparing the word descriptors of the first and third rules and determining that the first rule'"'"'s word descriptors are more specific than the third rule'"'"'s word descriptors; and

selecting to disambiguate the semantically ambiguous word using the first rule rather than the third rule based on the determination that the first rule'"'"'s word descriptor are more specific.

8. A machine for semantically disambiguating words;

the machine comprising;

an input text;

a set of rules for use in semantically disambiguating words, the set of rules including rules derived from two or more types of information in a corpus, each of the rules derived from at least two of the types of corpus information being applicable to words occurring in specified context; and

a processor connected for accessing the input text and the set of rules;

the processor, in using the set of rules to disambiguate words in the input text, operating to;

obtain context information about the context in which a semantically ambiguous word occurs in the input text;

a first rule derived from a first type of corpus information and a second rule derived from a second type of corpus information both being applicable to words occurring in context; and

select the first rule rather than the second rule to disambiguate the semantically ambiguous word according to a selection order based on the types of corpus information from which the rules are derived.

9. A stored rule set for use by a processor in semantically disambiguating words in text;

the stored rule set comprising;

a storage medium; and

rules data stored by the storage medium;

the rules data being accessible by the processor to obtain a set of semantic disambiguation rules derived from information in a corpus, the rules being useable by the processor in disambiguating a semantically ambiguous word that occurs in a context in an input text;

the corpus including two or more types of information, rules derived from at least first and second types of corpus information being applicable to words occurring in specified contexts, a first rule derived from the first type of corpus information and a second rule derived from the second type of corpus information both being applicable to words occurring in the context in which the semantically ambiguous word occurs in the input text;

the rules data further being accessible by the processor to obtain, for each of a set of the rules, a type of corpus information from which the rule was derived, for use in selecting the first rule rather than the second rule to disambiguate the semantically ambiguous word according to a selection order based on the types of corpus information from which the rules are derived.

10. The stored rule set of claim 9 in which the rules data include an item of data for each rule in the set;

each rule'"'"'s item of data being accessible by the processor to obtain the type of corpus information from which the rule was derived.

11. The stored rule set of claim 10 in which each rule'"'"'s item of data can further be accessed by the processor to obtain a type of relationship, a set of descriptors of words to which the relationship can apply, and a disambiguated sense of one of the words to which the relationship can apply.

12. The stored rule set of claim 11 in which, for at least one of the rules in the set, the set of descriptors includes a word and an ambiguity class applicable to another word.

13. The stored rule set of claim 9 in which the corpus is a dictionary and the types of corpus information include two or more of collocates, idioms, compounds, structure examples, phrasal verb examples, usages, and general examples.

14. A method of operating a first machine to transfer data to a second machine over a network, the second machine including a memory, an input text, and a processor connected for accessing the memory and for accessing the input text;

the method comprising;

establishing a connection between the first and second machines over the network; and

operating the first machine to transfer rules data to the memory of the second machine;

the rules data being accessible by the processor to obtain a set of semantic disambiguation rules derived from information in a corpus, the rules being useable by the processor in disambiguating a semantically ambiguous word that occurs in a context in the input text;

the corpus including two or more types of information, rules derived from at least first and second types of corpus information being applicable to words occurring in specified contexts, a first rule derived from the first type of corpus information and a second rule derived from the second type of corpus information both being applicable to words occurring in the context in which the semantically ambiguous words occurs in the input text;

the rules data further being accessible by the processor to obtain, for each of a set of rules, a type of corpus information from which the rule was derived, for use in selecting the first rule rather than the second rule to disambiguate the semantically ambiguous word according to a selection order based on the types of corpus information from which the rules are derived.

15. A method of semantically disambiguating words using rules, the rules including rules derived from two or more types of information in a corpus, rules derived from at least two of the types being applicable to words occurring in specified contexts;

the method comprising;

(A) obtaining context information about a context in which a semantically ambiguous word occurs in an input text;

a first rule derived from a first type of corpus information and a second rule derived from a second type of corpus information both being applicable to words occurring in the context; and

(B) first selecting the first rule rather than the second rule to disambiguate the semantically ambiguous word according to a selection order based on the types of corpus information from which the rules are derived, but where the first selecting of the first rule according to a selection order based on the types of corpus information is not applicable for selection, then second selecting the first rule rather than the second rule to disambiguate the semantically ambiguous word according to a distance-based selection rule for selection.