Natural language processing of disfluent sentences
First Claim
Patent Images
1. A computer-implemented method for processing spoken language comprising:
- converting spoken words into a text word sequence in a processor-based natural language processing system executing program code;
tagging words in the text word sequence with part-of-speech (POS) tags through a part-of-speech tagger component of the system; and
tagging edited words in the text word sequence using a disfluence identifier component of the system that operates with a feature set created with techniques comprising;
matching only the highest level POS tags in a multi-level hierarchy of such tags, wherein the highest level of the hierarchy comprises categories of tags including a noun category, a verb category, an adjective category, and an adverb category;
processing a resulting sequence of word-POS-tag pairs to mark each word in a text sequence with an edited-word-tag;
removing sequence-related errors in edited-word-tag information before parsing the text word sequence;
parsing the text word sequence into machine instructions with the aid of POS-tag and edited-word-tag information; and
allowing single mismatches in POS-tag sequences of rough copy, wherein rough copy in a string of POS-tagged words produces candidates for any potential pairs of reparanda and repairs by applying an algorithm to the string of POS-tagged words.
1 Assignment
0 Petitions
Accused Products
Abstract
An advanced model that includes new processes is provided for use as a component of an effective disfluency identifier. The disfluency identifier tags edited words in transcribed speech. A speech recognition unit in combination with a part-of-speech tagger, a disfluency identifier, and a parser form a natural language system that helps machines properly interpret spoken utterances.
-
Citations
14 Claims
-
1. A computer-implemented method for processing spoken language comprising:
-
converting spoken words into a text word sequence in a processor-based natural language processing system executing program code; tagging words in the text word sequence with part-of-speech (POS) tags through a part-of-speech tagger component of the system; and tagging edited words in the text word sequence using a disfluence identifier component of the system that operates with a feature set created with techniques comprising; matching only the highest level POS tags in a multi-level hierarchy of such tags, wherein the highest level of the hierarchy comprises categories of tags including a noun category, a verb category, an adjective category, and an adverb category; processing a resulting sequence of word-POS-tag pairs to mark each word in a text sequence with an edited-word-tag; removing sequence-related errors in edited-word-tag information before parsing the text word sequence; parsing the text word sequence into machine instructions with the aid of POS-tag and edited-word-tag information; and allowing single mismatches in POS-tag sequences of rough copy, wherein rough copy in a string of POS-tagged words produces candidates for any potential pairs of reparanda and repairs by applying an algorithm to the string of POS-tagged words. - View Dependent Claims (2, 3, 4, 5)
-
-
6. A processor-based system for processing spoken language comprising:
-
a speech recognition unit of the processor-based system configured to execute program code to convert spoken words into a text word sequence; a part-of-speech (POS) tagger of the processor-based system configured to execute program code to tag words in the text word sequence with part-of-speech tags; a disfluence identifier of the processor-based system configured to execute program code to tag edited words in the text word sequence with a feature set created with techniques comprising, matching only the highest level POS tags in a multi-level hierarchy of such tags; a parser for parsing the text word sequence into machine instructions with the aid of POS-tag and edited-word-tag information; and modifying a conventional definition of a rough copy by allowing single mismatching in POS-tag sequences of rough copy, wherein the conventional definition of rough copy in a string of tagged words has the form of ∂
1β
λ
∂
2, where,∂
1 (the source) and ∂
2 (the copy) both begin with non-punctuation;the strings of non-punctuation POS tags of ∂
1 and ∂
2 are identical;β
(the free final) consists of zero or more sequences of a free final word (see below) followed by optional punctuation; andinterregnum “
λ
”
consists of sequences of an interregnum string followed by optional punctuation. - View Dependent Claims (7, 8, 9, 10, 11)
-
-
12. A computer-implemented method for creating a disfluence identifier model comprising:
-
analyzing the distribution of speech repairs in transcribed speech in a processor-based natural language processing system executing program code; designating major categories of tags in a multi-level hierarchy of tags in a part-of-speech tagger unit of the natural language processing system, wherein the major categories comprise a noun category, a verb category, an adjective category, and an adverb category; choosing conditioning variables commensurate with the distribution of speech repairs; using a rough copy identifier with the conditioning variables to generate a feature set, in a disfluence identifier unit of the natural language processing system, wherein the rough copy identifier allows single mismatches in POS-tag sequences of rough copy, and wherein POS-tags are subordinate categories to the designated major categories in the multi-level hierarchy; and
,weighting the feature set according to an iterative algorithm run on training data and executed in the natural language processing system, wherein the rough copy identifier allows single mismatches in POS-tag sequences of rough copy. - View Dependent Claims (13, 14)
-
Specification