# US 12,217,737 B2 · Methods and apparatus for correcting failures in automated speech recognition systems > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,217,737 B2 - Document ID: US12217737B2 - Status: Granted - Kind: B2 - Application number: 18/581,666 - Filed: 2024-02-20 - Published: 2024-06-13 - Granted: 2025-02-04 - Earliest claimed: 2020-05-04 - Assignees: Adeia Guides Inc. - Applicants: Adeia Guides Inc. - Inventors: Ankur Anil Aher; Jeffry Copps Robert Jose - Web page: https://patentgrep.com/patent/US12217737B2 ## CPC - G10L 15/01: Assessment or evaluation of speech recognition systems - G06F 16/90332: {Natural language query formulation or dialogue systems} - G10L 15/063: {Training} - G10L 15/075: {supervised, i.e. under machine guidance} - G10L 15/1815: {Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning} - G10L 15/20: Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech (G10L21/02 takes precedence) - G10L 15/32: Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract Systems and methods are disclosed and described for correcting errors in ASR transcriptions. For an incorrect transcription, different words or phrases from the transcription, and/or related words or phrases, are submitted as hint words to the ASR system, and the voice query is submitted again, to determine new transcriptions. This process is repeated with different transcription terms, until a different and more proper transcription is generated. This increases the accuracy of ASR systems. ## Claims ### Claim 1 A method of correcting errors in searches for content resulting from incorrect transcriptions by an automated speech recognition (ASR) system, the method comprising: receiving an incorrect transcription output by an ASR module, the incorrect transcription comprising a plurality of transcription terms; successively submitting different subsets of the plurality of transcription terms with one or more hint words to the ASR module; receiving from the ASR module a plurality of resulting transcriptions, based on the successively submitted different subsets of the plurality of transcription terms with the one or more hint words, that are each different from the incorrect transcription; determining a most common transcription from the received plurality of resulting transcriptions; submitting the most common transcription as a search query of a database; and providing for display one or more representations of content returned by the search query of the database using the most common transcription. ### Claim 2 (depends on 1) The method of claim 1, wherein incorrect search results from the search query of the database are based on the incorrect transcription output. ### Claim 3 (depends on 1) The method of claim 1, wherein a transcription term of the plurality of transcription terms is assigned a popularity score, and wherein a low popularity score indicates that the transcription term is not often used by the search query of the database. ### Claim 4 (depends on 3) The method of claim 3, further comprising: comparing the popularity score associated with a particular transcription term of the plurality of transcription terms to a predetermined popularity score value; determining, based on the comparing, that the popularity score associated with the particular transcription term is less than the predetermined popularity score value; and determining that the particular transcription term is uncommon. ### Claim 5 (depends on 1) The method of claim 1, further comprising determining an occurrence of an improper search by determining one or more of: whether no content was selected by the search query of the database as a result of a search; whether the content returned by the search query of the database as the result of the search has an associated relevance score less than a predetermined value; whether terms of a transcription are not connected or weakly connected in a knowledge graph of terms of the content; or whether the terms of the transcription have popularity scores below a predetermined value. ### Claim 6 (depends on 5) The method of claim 5, wherein the determining that the occurrence of the improper search resulted from the incorrect transcription further comprises determining a presence of the incorrect transcription according to one or more of: whether a transcription score of the transcription corresponding to the improper search is less than a predetermined value; whether one or more terms of the transcription corresponding to the improper search have a homonym; whether the one or more terms of the transcription corresponding to the improper search have a corresponding pronunciation difficulty metric greater than a predetermined value; a noise level of a query resulting in the transcription corresponding to the improper search; whether the one or more terms of the transcription corresponding to the improper search is a mispronounced term; or whether the transcription corresponding to the improper search comprises more than one term. ### Claim 7 (depends on 1) The method of claim 1, further comprising: receiving from the ASR module one resulting transcription, based on the successively submitted different subsets of the plurality of transcription terms with the one or more hint words, that is different from the incorrect transcription; submitting the one resulting transcription to the search query of the database; and providing for display representations of the content returned by the search query of the database using the one resulting transcription. ### Claim 8 (depends on 1) The method of claim 1, further comprising, if the search query of the database generated an occurrence of an improper search and if the occurrence of the improper search did not result from the incorrect transcription, generating an error message and providing for display the one or more representations of the content returned by the search query of the database. ### Claim 9 (depends on 1) The method of claim 1, further comprising, if none of the different subsets of the plurality of transcription terms with the one or more hint words results in a resulting transcription that is different from the incorrect transcription, generating an error message and providing for display the one or more representations of the content returned by the search query of the database. ### Claim 10 (depends on 1) The method of claim 1, further comprising: conditionally performing the successively submitting if the search query of the database generated an occurrence of an improper search and if the occurrence of the improper search resulted from the incorrect transcription. ### Claim 11 A system of correcting errors in searches for content resulting from incorrect transcriptions by an automated speech recognition (ASR) system, the system comprising: a control circuitry; and an input/output (I/O) circuitry configured to; receive an incorrect transcription output by an ASR module, the incorrect transcription comprising a plurality of transcription terms; successively submit different subsets of the plurality of transcription terms with one or more hint words to the ASR module; receive from the ASR module a plurality of resulting transcriptions, based on the successively submitted different subsets of the plurality of transcription terms with the one or more hint words, that are each different from the incorrect transcription; determine, using the control circuitry, a most common transcription from the received plurality of resulting transcriptions; submit the most common transcription as a search query of a database; and provide for display one or more representations of content returned by the search query of the database using the most common transcription. ### Claim 12 (depends on 11) The system of claim 11, wherein incorrect search results from the search query of the database are based on the incorrect transcription output. ### Claim 13 (depends on 11) The system of claim 11, wherein a transcription term of the plurality of transcription terms is assigned a popularity score, and wherein a low popularity score indicates that the transcription term is not often used by the search query of the database. ### Claim 14 (depends on 13) The system of claim 13, wherein the control circuitry is further configured to: compare the popularity score associated with a particular transcription term of the plurality of transcription terms to a predetermined popularity score value; determine, based on the comparing, that the popularity score associated with the particular transcription term is less than the predetermined popularity score value; and determine that the particular transcription term is uncommon. ### Claim 15 (depends on 11) The system of claim 11, wherein the control circuitry is further configured to determine an occurrence of an improper search by determining one or more of: whether no content was selected by the search query of the database as a result of a search; whether the content returned by the search query of the database as the result of the search has an associated relevance score less than a predetermined value; whether terms of a transcription are not connected or weakly connected in a knowledge graph of terms of the content; or whether the terms of the transcription have popularity scores below a predetermined value. ### Claim 16 (depends on 15) The system of claim 15, wherein the control circuitry is further configured to determine that the occurrence of the improper search resulted from the incorrect transcription by determining a presence of the incorrect transcription according to one or more of: whether a transcription score of the transcription corresponding to the improper search is less than a predetermined value; whether one or more terms of the transcription corresponding to the improper search have a homonym; whether the one or more terms of the transcription corresponding to the improper search have a corresponding pronunciation difficulty metric greater than a predetermined value; a noise level of a query resulting in the transcription corresponding to the improper search; whether the one or more terms of the transcription corresponding to the improper search is a mispronounced term; or whether the transcription corresponding to the improper search comprises more than one term. ### Claim 17 (depends on 11) The system of claim 11, wherein the I/O circuitry is further configured to: receive from the ASR module one resulting transcription, based on the successively submitted different subsets of the plurality of transcription terms with the one or more hint words, that is different from the incorrect transcription; submit the one resulting transcription to the search query of the database; and provide for display representations of the content returned by the search query of the database from the one resulting transcription. ### Claim 18 (depends on 11) The system of claim 11, wherein the I/O circuitry is further configured to: if the search query of the database generated an occurrence of an improper search and if the occurrence of the improper search did not result from the incorrect transcription, generate an error message and provide for display the one or more representations of the content returned by the search query of the database. ### Claim 19 (depends on 11) The system of claim 11, wherein the I/O circuitry is further configured to: if none of the different subsets of the plurality of transcription terms with the one or more hint words results in a resulting transcription that is different from the incorrect transcription, generate an error message and providing for display the one or more representations of the content returned by the search query of the database. ### Claim 20 (depends on 11) The system of claim 11, wherein the I/O circuitry is further configured to: conditionally perform the successively submitting if the search query of the database generated an occurrence of an improper search and if the occurrence of the improper search resulted from the incorrect transcription.