LANGUAGE PROCESSOR

• US 20100241421A1
• Filed: 03/15/2010
• Published: 09/23/2010
• Est. Priority Date: 03/20/2009
• Status: Active Grant

• Current Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Original Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Inventors: FUNAKOSHI, Kotaro
• Granted Patent: US 8,660,837 B2
• Time in Patent Office: Days
• US Class Current: 704/201
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning

Abstract

A language processor according to the present invention includes a probability calculating section (103) having a probabilistic model by which a probability that a set of partial referring expression pairs represents an object, each pair consisting of an attribute value expression and a part expression modified by the attribute value expression, the set of referring expression pairs including a pair concerning the whole object, is represented by a product of probabilities each of which is a probability that each partial referring expression pair represents the object. The probability that each partial referring expression pair represents the object is obtained using a probability that the object is chosen, a probability that represents the salience of a part of the object, and a probability representing relevance between the part of the object and an attribute value, a probability that the part expression is used for a part of the object, and a probability that the attribute value expression is used for an attribute value. It further includes at least one memory section (101) for storing values of probabilities used for the probabilistic model. The probability calculating section calculates the probability that the set of partial referring expression pairs represents the object using the values of probabilities stored in the at least one memory section and the probabilistic model.

Claims

1. A language processor using a probabilistic model, comprising:

a probability calculating section having the probabilistic model by which a probability that a set of partial referring expression pairs represents an object, each pair consisting of an attribute value expression and a part expression modified by the attribute value expression, the set of referring expression pairs including a pair concerning the whole object, is represented by a product of probabilities each of which is a probability that each partial referring expression pair represents the object, the probability that each partial referring expression pair represents the object is obtained using a probability that the object is chosen, a salience probability that represents the salience of a part of the object, and a probability representing relevance between the part of the object and an attribute value, a probability of using the part expression that the part expression is used for a part of the object, and a probability of using the attribute value expression that the attribute value expression is used for an attribute value;

at least one memory section for storing values of salience probabilities of parts of respective objects, values of probabilities representing relevance between parts of respective objects and attribute values, values of probabilities of using part expressions and values of probabilities of using attribute value expressions,

wherein the probability calculating section calculates the probability that the set of partial referring expression pairs represents the object using values of probabilities stored in the at least one memory section and the probabilistic model.

3. Language processor according to claim 1, wherein the probability that the object is chosen can be changed according to the situation.

4. A language processor according to claim 1, wherein values of probabilities of using part expressions and values of probabilities of using attribute value expressions are determined commonly for respective objects.

5. A language processor according to claim 1, wherein values of probabilities of using part expressions and values of probabilities of using attribute value expressions are determined individually for respective kinds of objects.

6. A language processor according to claim 1, wherein values of probabilities of using part expressions and values of probabilities of using attribute value expressions are determined individually for respective objects.

7. A language processor according to claim 1, further comprising an object identifying section which when identifying the object represented by the set of partial referring expression pairs among potential objects for which probabilities that the potential objects are represented by the set of partial referring expression pairs are obtained using the probability calculating section, identifies the object with the maximum probability among the potential objects as the object represented by the set of partial referring expression pairs.

8. A language processor according to claim 7, further comprising a structure analyzing section which generates at least one set of partial referring expression pairs from natural language.

9. A language processor according to claim 8, wherein the structure analyzing section generates plural sets of partial referring expression pairs as potential sets of partial referring expression pairs and the object identifying section identifies the object for which the probability that one of the plural sets of partial referring expression pairs represents the object is maximum, as the object that the one of the sets represents.

10. A language processor according to claim 1, further comprising a referring expression selecting section which when selecting a set of partial referring expression pairs that represents an object among potential sets of partial referring expression pairs for which probabilities that the potential sets of partial referring expression pairs represent the object are obtained using the probability calculating section, selects the set of partial referring expression pairs for which the probability is maximum as the set of partial referring expression pairs representing the object.

11. A language processor according to claim 10, further comprising a surface expression realizing section which generates a natural language expression from the set of partial referring expression pairs selected by the referring expression selecting section.

2. A language processor using a probabilistic model, comprising:

a probability calculating section having the probabilistic model by which a probability

Pr(o|e)

that an object is represented by a set of partial referring expression pairs,

e={(e₁^p,e₁^a),(e₂^p,e₂^a), . . . , (e_n^p,e_n^a)}

each pair (e_i^p,e_i^a) consisting of an attribute value expression and a part expression modified by the attribute value expression, the set of partial referring expression pairs including a pair concerning the whole object, is represented by a product of probabilities each of which is a probability

Pr(o|e_i^p,e_i^a)

each partial referring expression pair represents the object, the probability that each partial referring expression pair represents the object is represented by a product of a probability

Pr(o|e)

that the object is chosen, a product of a probability of using the part expression of the pair)

Pr(e_i^p|p_u,o)

that the part expression e_i^p is used for a part p_u of the object o, a probability of using the attribute value expression of the pair

Pr(e_i^a|a_v,o)

that the attribute value expression e_i^a is used for an attribute value a_v, a salience probability)

Pr(p_u|o)

that represents the salience of the part of the object, and a fitness probability)

Pr(a_v|p_u,o)

that represents the degree of fitness of the part of the object to the attribute value;

at least one memory section for storing values of salience probabilities of parts of respective objects, values of fitness probabilities, values of probabilities of using part expressions and values of probabilities of using attribute value expressions,

wherein the probability calculating section calculates the probability that the set of partial referring expression pairs represents the object using values of probabilities stored in the at least one memory section and the probabilistic model.

12. A probability calculating method by which a probability that a set of partial referring expression pairs represents an object, each pair consisting of an attribute value expression and a part expression modified by the attribute value expression, the set of referring expression pairs including a pair concerning the whole object, is obtained in a language processor provided with a probability calculating section and at least one memory section, the method comprising the steps of

obtaining, by the probability calculating section, a probability that each partial referring expression pair represents the object using a probability that the object is chosen, a salience probability that represents the salience of a part of the object, and a probability representing relevance between the part of the object and an attribute value, a probability of using the part expression that the part expression is used for a part of the object, and a probability of using the attribute value expression that the attribute value expression is used for an attribute value, values of which are stored in the at least one memory section;

and

obtaining, by the probability calculating section, the probability that the set of partial referring expression pairs represents the object by obtaining a product of probabilities, each of which is the probability that each partial referring expression pair represents the object.