Transcription correction using multi-token structures

Patent metadata

Field	Value
Publication number	US-2016275071-A1
Application number	US-201615171149-A
Country	US
Kind code	A1
Filing date	Jun 2, 2016
Priority date	Jan 27, 2015
Publication date	Sep 22, 2016
Grant date	—

Abstract

Examples of the present disclosure describe generation of a multi-arc confusion network to improve, for example, an ability to return alternatives to output generated. A confusion network comprising token representations of lexicalized hypotheses and normalized hypotheses is generated. Each arc of the confusion network represents a token of a lexicalized hypothesis or a normalized hypothesis. The confusion network is transformed into a multi-arc confusion network, wherein the transforming comprising realigning at least one token of the confusion network to span multiple arcs of the confusion network. Other examples are also described.

First claim

1 - 20 . (canceled)

21 . A method comprising:
receiving, through a computing device executing an input understanding application, an input requesting an alternative to a result that is displayed through the input understanding application; accessing a multi-arc confusion network to identify one or more alternatives for the result; outputting, from the multi-arc confusion network, the one or more alternatives for the result through the input understanding application; and displaying, on a display connected with the computing device, the one or more alternatives.

22 . The method according to claim 21 , wherein the multi-arc confusion network comprises an arc identifying a token representation for the result, and wherein the arc spans multiple nodes of the multi-arc confusion network.

23 . The method according to claim 22 , wherein the one or more alternatives are identified from the multiple nodes of the arc associated with the result.

24 . The method according to claim 21 , wherein the one or more alternatives are output based on scores associated with the one or more alternatives generated from processing associated with the multi-arc confusion network.

25 . The method according to claim 21 , wherein the input understanding application is an intelligent personal assistant.

26 . The method according to claim 21 , further comprising receiving an utterance, generating a confusion network comprising token representations of normalized hypotheses for the utterance, wherein an arc of the confusion network represents a token of a normalized hypothesis, and transforming the confusion network into the multi-arc confusion network by realigning at least one token of the confusion network to span multiples arcs of the confusion network.

27 . A system comprising:
at least one computer processor; a display; and a memory operatively connected with the at least one computer processor, wherein the memory stores computer-executable instructions that, when executed by the at least one computer processor, causes the at least one computer processor to execute a method that comprises:
receiving, through a computing device executing an input understanding application, an input requesting an alternative to a result that is displayed through the input understanding application;
accessing a multi-arc confusion network to identify one or more alternatives for the result;
outputting, from the multi-arc confusion network, the one or more alternatives for the result through the input understanding application; and
displaying, on the display, the one or more alternatives.

28 . The system according to claim 27 , wherein the multi-arc confusion network comprises an arc identifying a token representation for the result, and wherein the arc spans multiple nodes of the multi-arc confusion network.

29 . The system according to claim 28 , wherein the one or more alternatives are identified from the multiple nodes of the arc associated with the result.

30 . The system according to claim 27 , wherein the one or more alternatives are output based on scores associated with the one or more alternatives generated from processing associated with the multi-arc confusion network.

31 . The system according to claim 27 , wherein the input understanding application is an intelligent personal assistant.

32 . The system according to claim 27 , wherein the method further comprises:
receiving an utterance, generating a confusion network comprising token representations of normalized hypotheses for the utterance, wherein an arc of the confusion network represents a token of a normalized hypothesis, and transforming the confusion network into the multi-arc confusion network by realigning at least one token of the confusion network to span multiples arcs of the confusion network.

33 . A system comprising:
at least one computer processor; and a memory operatively connected with the at least one computer processor, wherein the memory stores computer-executable instructions that, when executed by the at least one computer processor, causes the at least one computer processor to execute a method that comprises:
receiving data associated with an utterance;
generating a confusion network comprising token representations of normalized hypotheses in response to the received utterance, wherein each arc of the confusion network represents a token of a normalized hypothesis;
transforming the generated confusion network into a multi-arc confusion network, wherein the transforming comprising realigning at least one token of the confusion network to span multiples arcs of the confusion network; and
outputting one or more lexical results for the utterance based on processing using the multi-arc confusion network.

34 . The system according to claim 33 , wherein the outputting further comprises outputting two lexical results for the utterance.

35 . The system according to claim 34 , wherein the two lexical results comprise a lexical hypothesis for the utterance and an alternative lexical hypothesis for the utterance.

36 . The system according to claim 33 , wherein the system is a computing device associated with an input understanding service.

37 . The system according to claim 36 , wherein the method further comprises transmitting the one or more lexical results to a processing device connected with the input understanding service over a distributed network.

38 . The system according to claim 33 , wherein the realigning further comprises changing a starting point and an ending point for an arc associated with the token to span multiple arcs representing elements of a normalized hypothesis.

39 . The system according to claim 33 , wherein the method further comprises receiving an input requesting an alternative to the one or more lexical results, accessing the multi-arc confusion network to identify one or more alternatives for the one or more lexical results, and outputting, using the multi-arc confusion network, the one or more alternatives to a processing display for display of the one or more alternatives.

40 . The system according to claim 33 , wherein the outputting further comprises displaying the one or more lexical results using a display connected with the system.

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Levit Michael
• Ozertem Umut
• Parthasarathy Sarangarajan
• Varadharajan Padma
• Raghunathan Karthik
• Alphonso Issac

Classifications

• G10L15/187 Primary

  Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams · CPC title

• G10L15/083

  Recognition networks (G10L15/142, G10L15/16 take precedence) · CPC title

• G10L15/197

  Probabilistic grammars, e.g. word n-grams · CPC title

• G10L15/02

  Feature extraction for speech recognition; Selection of recognition unit · CPC title

• G10L15/22

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title