Method and apparatus for adapting a class entity dictionary used with language models

• US 7,124,080 B2
• Filed: 11/13/2001
• Issued: 10/17/2006
• Est. Priority Date: 11/13/2001
• Status: Expired due to Fees

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Zhang, Feng, Li, Mingjing, Chen, Zheng, Gao, Jianfeng
• Primary Examiner(s): Hudspeth, David
• Assistant Examiner(s): Sked, Matthew J.
• Application Number:
  US10/008,432
• Publication Number:
  US 20030093263A1
• Time in Patent Office: 1,799 Days
• Field of Search: None
• US Class Current: 704/244
• CPC Class Codes: G06F 40/216 using statistical methods
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking

Abstract

A method and apparatus are provided for augmenting a language model with a class entity dictionary based on corrections made by a user. Under the method and apparatus, a user corrects an output that is based in part on the language model by replacing an output segment with a correct segment. The correct segment is added to a class of segments in the class entity dictionary and a probability of the correct segment given the class is estimated based on an n-gram probability associated with the output segment and an n-gram probability associated with the class. This estimated probability is then used to generate further outputs.

Claims

1. A method of decoding input, the method comprising:

identifying possible sequences of words from the input;

using a class-based language model and a class entity dictionary to select one of the possible sequences of words as an output sequence, the class entity dictionary providing the probability of an entity given a class;

receiving modifications made to the output sequence by receiving a modified entity that represents a modification of a decoded entity in the ouput sequence; and

using the modifications to change the class entity dictionary by adding the modified entity to the class entity dictionary and estimating a probability for the modified entity given the class to which the entity is added based on an n-gram probability associated with the decoded entity and an n-gram probability associated with the class to which the modified entity is added.

2. The method of claim 1 wherein receiving modifications comprises receiving a modified entity that represents a modification of a decoded entity in the output sequence and wherein the modified entity is not found in the class entity dictionary but the decoded entity is found in the class entity dictionary.

3. The method of claim 2 further comprising decreasing the probability of the decoded entity.

4. A computer-readable medium having computer-executable instructions for performing steps comprising:

generating a sequence of words based in part on a class entity dictionary that provides probabilities for entities in at least one class;

receiving a modification to the sequence of words such that a decoded entity in the sequence of words is modified into a modified entity; and

adding the modified entity to the class entity dictionary and estimating a probability for the modified entity in the class entity dictionary based on an n-gram probability associated with the decoded entity and an n-gram probability associated with the class, the probability providing the probability of the modified entity given a class.

5. The computer-readable medium of claim 4 further comprising decreasing the probability of an entity.

6. The computer-readable medium of claim 5 wherein decreasing a probability further comprises:

determining that the decoded entity is in the class entity dictionary; and

decreasing the probability of the decoded entity.