David N Weise

US Patent:

6952666, Oct 4, 2005

Filed:

Jul 20, 2000

Appl. No.:

09/620745

Inventors:

David N. Weise - Kirkland WA, US

Assignee:

Microsoft Corporation - Redmond WA

International Classification:

G10L015/00

US Classification:

704 9, 704 10

Abstract:

A natural language parse ranker of a natural language processing (NLP) system employs a goodness function to rank the possible grammatically valid parses of an utterance. The goodness function generates a statistical goodness measure (SGM) for each valid parse. The parse ranker orders the parses based upon their SGM values. It presents the parse with the greatest SGM values as the one that most likely represents the intended meaning of the speaker. The goodness function of this parse ranker is highly accurate in representing the intended meaning of a speaker. It also has reasonable training data requirements. With the parse ranker, the SGM of a particular parse is the combination of all of the probabilities of each node within the parse tree of such parse. The probability at a given node is the probability of taking a transition (“grammar rule”) at that point. The probability at a node is conditioned on highly predicative linguistic phenomena, such as “phrase levels,” “null transitions,” and “syntactic history”.

US Patent:

6985851, Jan 10, 2006

Filed:

Jul 17, 2001

Appl. No.:

09/907315

Inventors:

David Neal Weise - Kirkland WA, US
Aravind Bala - Seattle WA, US

Assignee:

Microsoft Corporation - Redmond WA

International Classification:

G06F 17/27
G06F 17/28
G06F 17/21
G06F 17/20

US Classification:

704 9, 704 1, 704 4, 704 5, 704 10

Abstract:

A method of calculating trigram path probabilities for an input string of text containing a multi-word-entry (MWE) or a factoid includes tokenizing the input string to create a plurality of parse leaf units (PLUs). A PosColumn is constructed for each word, MWE, factoid and character in the input string of text which has a unique first (Ft) and last (Lt) token pair. TrigramColumns are constructed which define corresponding TrigramNodes each representing a trigram for three PosColumns. Forward and backward trigram path probabilities are calculated for each separate TrigramNode. The sums of all trigram path probabilities through each PLU are then calculated as a function of the forward and backward trigram path probabilities. Systems and computer-readable medium configured to implement the methods are also provided.

US Patent:

7003444, Feb 21, 2006

Filed:

Jul 12, 2001

Appl. No.:

09/904232

Inventors:

David Neal Weise - Kirkland WA, US

Assignee:

Microsoft Corporation - Redmond WA

International Classification:

G06F 17/27

US Classification:

704 9, 704257, 704 1, 704 10, 715533

Abstract:

A method and grammar checking system are provided that generate a stochastic score, or a statistical goodness measure, for each of an input string of text and one or more alternative strings of text. An alternative generator generates the alternative strings of text, and a ranking parser produces parse trees and corresponding statistical goodness measures for each of the strings. The string of text having the highest goodness measure is selected for recommendation to a user.

US Patent:

7003445, Feb 21, 2006

Filed:

Jul 20, 2001

Appl. No.:

09/909530

Inventors:

Kevin W. Humphreys - Redmond WA, US
David Neal Weise - Kirkland WA, US
Michael V. Calcagno - Kirkland WA, US

Assignee:

Microsoft Corporation - Redmond WA

International Classification:

G06F 17/28
G06F 17/20
G06F 17/27

US Classification:

704 9, 704 7, 704 8

Abstract:

A method of, and system for, generating a sentence from a semantic representation maps the semantic representation to an unordered set of syntactic nodes. Simplified generation grammar rules and statistical goodness measure values from a corresponding analysis grammar are then used to create a tree structure to order the syntactic nodes. The sentence is then generated from the tree structure. The generation grammar is a simplified (context free) version of a corresponding full (context sensitive) analysis grammar. In the generation grammar, conditions on each rule are ignored except those directly related to the semantic representation. The statistical goodness measure values, which are calculated through an analysis training phase in which a corpus of example sentences is processed using the full analysis grammar, are used to guide the generation choice to prefer substructures most commonly found in a particular syntactic/semantic context during analysis.

US Patent:

7124074, Oct 17, 2006

Filed:

Jun 14, 2005

Appl. No.:

11/151953

Inventors:

David Neal Weise - Kirkland WA, US
Aravind Bala - Redmond WA, US

Assignee:

Microsoft Corporation - Redmond WA

International Classification:

G06F 17/27
G06F 17/21
G06F 17/28

US Classification:

704 9, 704 4, 704 10

Abstract:

A method of calculating trigram path probabilities for an input string of text containing a multi-word-entry (MWE) or a factoid includes tokenizing the input string to create a plurality of parse leaf units (PLUs). A PosColumn is constructed for each word, MWE, factoid and character in the input string of text which has a unique first (Ft) and last (Lt) token pair. TrigramColumns are constructed which define corresponding TrigramNodes each representing a trigram for three PosColumns. Forward and backward trigram path probabilities are calculated for each separate TrigramNode. The sums of all trigram path probabilities through each PLU are then calculated as a function of the forward and backward trigram path probabilities. Systems and computer-readable medium configured to implement the methods are also provided.

US Patent:

7143036, Nov 28, 2006

Filed:

Aug 30, 2004

Appl. No.:

10/929167

Inventors:

David N. Weise - Kirkland WA, US

Assignee:

Microsoft Corporation - Redmond WA

International Classification:

G10L 15/00

US Classification:

704245, 704257

Abstract:

A natural language parse ranker of a natural language processing (NLP) system employs a goodness function to rank the possible grammatically valid parses of an utterance. The goodness function generates a statistical goodness measure (SGM) for each valid parse. The parse ranker orders the parses based upon their SGM values. It presents the parse with the greatest SGM value as the one that most likely represents the intended meaning of the speaker. The goodness function of this parse ranker is highly accurate in representing the intended meaning of a speaker. It also has reasonable training data requirements. With this parse ranker, the SGM of a particular parse is the combination of all of the probabilities of each node within the parse tree of such parse. The probability at a given node is the probability of taking a transition (“grammar rule”) at that point. The probability at a node is conditioned on highly predicative linguistic phenomena, such as “phrase levels,” “null transitions,” and “syntactic history”.

US Patent:

7184950, Feb 27, 2007

Filed:

Jul 8, 2005

Appl. No.:

11/177129

Inventors:

David Neal Weise - Kirkland WA, US

Assignee:

Microsoft Corporation - Redmond WA

International Classification:

G06F 17/27

US Classification:

704 9, 704 1, 704 2, 704 5

Abstract:

A method and grammar checking system are provided that generate a stochastic score, or a statistical goodness measure, for each of an input string of text and one or more alternative strings of text. An alternative generator generates the alternative strings of text, and a ranking parser produces parse trees and corresponding statistical goodness measures for each of the strings. The string of text having the highest goodness measure is selected for recommendation to a user.

US Patent:

7266491, Sep 4, 2007

Filed:

Jun 14, 2005

Appl. No.:

11/152352

Inventors:

Kevin W. Humphreys - Redmond WA, US
David Neal Weise - Kirkland WA, US
Michael V. Calcagno - Kirkland WA, US

Assignee:

Microsoft Corporation - Redmond WA

International Classification:

G06F 17/27

US Classification:

704 4, 704 9

Abstract:

A method of, and system for, generating a sentence from a semantic representation maps the semantic representation to an unordered set of syntactic nodes. Simplified generation grammar rules and statistical goodness measure values from a corresponding analysis grammar are then used to create a tree structure to order the syntactic nodes. The sentence is then generated from the tree structure. The generation grammar is a simplified (context free) version of a corresponding full (context sensitive) analysis grammar. In the generation grammar, conditions on each rule are ignored except those directly related to the semantic representation. The statistical goodness measure values, which are calculated through an analysis training phase in which a corpus of example sentences is processed using the full analysis grammar, are used to guide the generation choice to prefer substructures most commonly found in a particular syntactic/semantic context during analysis.