DETAILED ACTION
This action is in response to the application filed on May 17th, 2024. Claims 1-20 are pending and have been examined.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Objections
Claims 12 and 17 objected to because of the following informalities:
Regarding claim 12, the 5th limitation has an erroneous semicolon part way through.
Regarding claim 17, the 3rd limitation does not have a semicolon at the end.
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 5, 8, and 11 rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 5 recites the limitation "the connectionist temporal classification model" in line 2. There is insufficient antecedent basis for this limitation in the claim.
Claim 8 recites the limitation "the plurality of filtered embeddings" in line 1. There is insufficient antecedent basis for this limitation in the claim. Based on the language in prior claims, the examiner believes this should read “the plurality of filtered encodings”.
Claim 11 recites the limitation "the machine-learning model" in line 4. There is insufficient antecedent basis for this limitation in the claim. Based on the language in prior claims, the examiner believes this should read “the machine-trained model”.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-4, 6-7, 12-15, and 17-19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (US Pat. No. 12,154,558 B1 hereinafter Wang), in view of Quan, Zong-Feng (CN Pat. No. 114596868 A hereinafter Quan).
Regarding claim 1, Wang discloses a computer-implemented method for recognizing speech, the method comprising: processing, by the processor, the plurality of filtered encodings to generate a plurality of audio embeddings (Wang, Col. 14, lines 22-25: "The ASR component 250 receives audio data 211 (for example, received from a local device 110 having processed audio detected by a microphone by an acoustic front end (AFE) or other component)."; lines 29-35: "The audio data 211 may be audio data that has been digitized (for example by an AFE) into frames representing time intervals for which the AFE determines a number of values, called features, representing the qualities of the audio data, along with a set of those values, called a feature vector, representing the features/qualities of the audio data within the frame."); mapping, by the processor, each audio embedding of the plurality of audio embeddings to a textual embedding using a speech adapter to generate a plurality of combined embeddings (Wang, Fig. 3, 250; Col. 9, lines 44-48: "The ASR component 250 transcribes audio data into one or more ASR hypotheses 305a and 305b (collectively “ASR hypotheses 305”) (e.g., one or more different textual or other representations of the speech contained in the audio data)."); receiving, by the processer, one or more specific textual embeddings from a domain-specific entity retriever based on the plurality of filtered encodings (Wang, Fig. 3, 270; Col. 10, lines 14-21: "In some implementations, the orchestrator component 230 can send the 1-best ASR hypotheses 305a to the ER hypothesis generator 270. The ER hypotheses generator 270, described in further detail below, can generate additional entity mention hypotheses based on the slot text in the 1-best ASR hypothesis 305a. The ER hypothesis generator 270 can return slot text 315b for one or more of its hypotheses to the orchestrator component 230."); providing, by the processer, the plurality of combined embeddings and the one or more specific textual embeddings to a machine-trained model; and receiving, by the processor, a textual output representing speech from the speech input from the machine-trained model (Wang, Fig. 3, 290; Col. 10, lines 36-44: "The orchestrator component 230 can forward the NLU results 210 and the entity data 320 to a skill 290 and/or a skill system 125 for processing. The orchestrator component 230 may receive output data 325 from the skill 290 and/or skill system 125 and cause performance of one or more actions in response to the input audio data 211. In some implementations, the orchestrator component 230 can forward the output data 325 to the TTS component 280 for generating a spoken-word response to the input audio data 211."). However, Wang fails to expressly recite performing, by a processor, blank filtering on a received speech input to generate a plurality of filtered encodings.
Quan teaches performing, by a processor, blank filtering on a received speech input to generate a plurality of filtered encodings (Quan, Page 4, step 130: "step 130, determining the effective frame from the characteristic frame sequence based on the blank probability. In order to reduce the calculation amount of the voice recognition the blank frame can be removed from the characteristic frame sequence, only keeping the effective frame, so as to shorten the characteristic frame sequence for subsequent calculation, so as to improve the calculation efficiency of the voice recognition Specifically, after calculating the blank probability of each characteristic frame, it can determine whether the characteristic frame is a valid frame based on the blank probability.").
Wang and Quan are analogous arts because they each belong to the same field of speech processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the entity resolution system of Wang to incorporate the teachings of Quan to filter blank frames from the audio input. Filtering blank frames reduces the amount of data that must be processed after filtering (Quan, Page 4, step 130). This improves the overall efficiency of the system.
Regarding claim 2, the rejection of claim 1 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. Quan further teaches wherein performing blank filtering comprises removing one or more frames from the speech input that do not include speech to generate the plurality of filtered encodings (Quan, Page 4, step 130: "step 130, determining the effective frame from the characteristic frame sequence based on the blank probability. In order to reduce the calculation amount of the voice recognition the blank frame can be removed from the characteristic frame sequence, only keeping the effective frame, so as to shorten the characteristic frame sequence for subsequent calculation, so as to improve the calculation efficiency of the voice recognition Specifically, after calculating the blank probability of each characteristic frame, it can determine whether the characteristic frame is a valid frame based on the blank probability."). The same motivation for claim 1 applies equally to claim 2.
Regarding claim 3, the rejection of claim 1 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. Quan further teaches wherein the plurality of filtered encodings are generated in part using a connectionist temporal classification model (Quan, Page 4, step 120: "step 120, calculating the blank probability of each characteristic frame in the characteristic frame sequence. The Connectionist Temporal Classification (CTC) is a method for avoiding input and output manual alignment."). The same motivation for claim 1 applies equally to claim 3.
Regarding claim 4, the rejection of claim 1 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. Wang further discloses wherein the speech adapter is trained using speech as an input and a predicted transcript as an output (Wang, Col. 11, line 65- Col. 12, line 1: "In some implementations, the language model 272 can be configured to calculate p(t′|previous_words) and p(t′|following_words) based on co-occurrence of the words in a training data set."; Col. 12, lines 5-10: “In an example operation, the ground truth of the entity should be “urbana free library,” which the ASR component 250 incorrectly transcribed as “banna free library.” The ER hypothesis generator 270 can generate candidates for “banna” based on its context word “free library” and then rerank them based on the phoneme similarity to “banna””).
Regarding claim 6, the rejection of claim 1 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. Wang further discloses wherein the domain-specific entity retriever is a dual encoder model that comprises keys and values, wherein the keys are acoustic encodings and the values are domain-specific entities (Wang, Col. 12, lines 36-42: "The ER component 266 may query a search index 262 and identify entity data 320 responsive to the audio data 211. The search index 262 may include one or more entity catalogs associated with the intent. The ER component 266 can return the entity data 320 including one or more entities matching the slot text 315.").
Regarding claim 7, the rejection of claim 6 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. Wang further discloses wherein the domain-specific entity retriever is trained using entities mentioned in a reference transcript of the speech input (Wang, Col. 23, lines 63-66: "The entity resolution component 266 and search index 262 can be generated by applying machine learning to a training data set, where the training set includes both lexical and phonetic information.").
Regarding claim 12, Wang discloses a computing system for recognizing speech, the computing system comprising: one or more processors (Wang, Col. 26, line 61- Col. 27, line 4: “Each of these devices (110/120/125) may include one or more controllers/processors (1304/1404), which may each include a central processing unit (CPU) for processing data and computer-readable instructions, and a memory (1306/1406) for storing data and instructions of the respective device.”); and a non-transitory, computer-readable medium comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations (Wang, Col. 27, lines 19-26: “Computer instructions for operating each device (110/120/125) and its various components may be executed by the respective device's controller(s)/processor(s) (1304/1404), using the memory (1306/1406) as temporary “working” storage at runtime. A device's computer instructions may be stored in a non-transitory manner in non-volatile memory (1306/1406), storage (1308/1408), or an external device(s).”), the operations comprising: processing the plurality of filtered encodings to generate a plurality of audio embeddings (Wang, Col. 14, lines 22-25: "The ASR component 250 receives audio data 211 (for example, received from a local device 110 having processed audio detected by a microphone by an acoustic front end (AFE) or other component)."; lines 29-35: "The audio data 211 may be audio data that has been digitized (for example by an AFE) into frames representing time intervals for which the AFE determines a number of values, called features, representing the qualities of the audio data, along with a set of those values, called a feature vector, representing the features/qualities of the audio data within the frame."); mapping each audio embedding of the plurality of audio embeddings to a textual embedding using a speech adapter; to generate a plurality of combined encodings (Wang, Fig. 3, 250; Col. 9, lines 44-48: "The ASR component 250 transcribes audio data into one or more ASR hypotheses 305a and 305b (collectively “ASR hypotheses 305”) (e.g., one or more different textual or other representations of the speech contained in the audio data)."); receiving one or more specific textual embeddings from a domain-specific entity retriever based on the plurality of filtered encodings (Wang, Fig. 3, 270; Col. 10, lines 14-21: "In some implementations, the orchestrator component 230 can send the 1-best ASR hypotheses 305a to the ER hypothesis generator 270. The ER hypotheses generator 270, described in further detail below, can generate additional entity mention hypotheses based on the slot text in the 1-best ASR hypothesis 305a. The ER hypothesis generator 270 can return slot text 315b for one or more of its hypotheses to the orchestrator component 230."); providing the plurality of combined embeddings and the one or more specific textual embeddings to a machine-trained model; and receiving a textual output representing speech from the speech input from the machine-trained model (Wang, Fig. 3, 290; Col. 10, lines 36-44: "The orchestrator component 230 can forward the NLU results 210 and the entity data 320 to a skill 290 and/or a skill system 125 for processing. The orchestrator component 230 may receive output data 325 from the skill 290 and/or skill system 125 and cause performance of one or more actions in response to the input audio data 211. In some implementations, the orchestrator component 230 can forward the output data 325 to the TTS component 280 for generating a spoken-word response to the input audio data 211."). However, Wang fails to expressly recite performing blank filtering on a received speech input to generate a plurality of filtered encodings.
Quan teaches performing blank filtering on a received speech input to generate a plurality of filtered encodings (Quan, Page 4, step 130: "step 130, determining the effective frame from the characteristic frame sequence based on the blank probability. In order to reduce the calculation amount of the voice recognition the blank frame can be removed from the characteristic frame sequence, only keeping the effective frame, so as to shorten the characteristic frame sequence for subsequent calculation, so as to improve the calculation efficiency of the voice recognition Specifically, after calculating the blank probability of each characteristic frame, it can determine whether the characteristic frame is a valid frame based on the blank probability.").
Wang and Quan are analogous arts because they each belong to the same field of speech processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the entity resolution system of Wang to incorporate the teachings of Quan to filter blank frames from the audio input. Filtering blank frames reduces the amount of data that must be processed after filtering (Quan, Page 4, step 130). This improves the overall efficiency of the system.
Regarding claim 13, the rejection of claim 12 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. Quan further teaches wherein performing blank filtering comprises removing one or more frames from the speech input that do not include speech to generate the plurality of filtered encodings (Quan, Page 4, step 130: "step 130, determining the effective frame from the characteristic frame sequence based on the blank probability. In order to reduce the calculation amount of the voice recognition the blank frame can be removed from the characteristic frame sequence, only keeping the effective frame, so as to shorten the characteristic frame sequence for subsequent calculation, so as to improve the calculation efficiency of the voice recognition Specifically, after calculating the blank probability of each characteristic frame, it can determine whether the characteristic frame is a valid frame based on the blank probability."). The same motivation for claim 12 applies equally to claim 13.
Regarding claim 14, the rejection of claim 12 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. Quan further teaches wherein the plurality of filtered encodings are generated in part using a connectionist temporal classification model (Quan, Page 4, step 120: "step 120, calculating the blank probability of each characteristic frame in the characteristic frame sequence. The Connectionist Temporal Classification (CTC) is a method for avoiding input and output manual alignment."). The same motivation for claim 12 applies equally to claim 14.
Regarding claim 15, the rejection of claim 12 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. Wang further discloses wherein the domain-specific entity retriever is a dual encoder model that comprises keys and values, wherein the keys are acoustic encodings and the values are domain-specific entities (Wang, Col. 12, lines 36-42: "The ER component 266 may query a search index 262 and identify entity data 320 responsive to the audio data 211. The search index 262 may include one or more entity catalogs associated with the intent. The ER component 266 can return the entity data 320 including one or more entities matching the slot text 315.").
Regarding claim 17, Wang discloses a non-transitory, computer-readable medium comprising instructions that, when executed by one or more processors, cause the one or more processors to perform operations (Wang, Col. 27, lines 19-26: “Computer instructions for operating each device (110/120/125) and its various components may be executed by the respective device's controller(s)/processor(s) (1304/1404), using the memory (1306/1406) as temporary “working” storage at runtime. A device's computer instructions may be stored in a non-transitory manner in non-volatile memory (1306/1406), storage (1308/1408), or an external device(s).”), the operations comprising: processing the plurality of filtered encodings to generate a plurality of audio embeddings (Wang, Col. 14, lines 22-25: "The ASR component 250 receives audio data 211 (for example, received from a local device 110 having processed audio detected by a microphone by an acoustic front end (AFE) or other component)."; lines 29-35: "The audio data 211 may be audio data that has been digitized (for example by an AFE) into frames representing time intervals for which the AFE determines a number of values, called features, representing the qualities of the audio data, along with a set of those values, called a feature vector, representing the features/qualities of the audio data within the frame."); mapping each audio embedding of the plurality of audio embeddings to a textual embedding using a speech adapter to generate a plurality of combined embeddings (Wang, Fig. 3, 250; Col. 9, lines 44-48: "The ASR component 250 transcribes audio data into one or more ASR hypotheses 305a and 305b (collectively “ASR hypotheses 305”) (e.g., one or more different textual or other representations of the speech contained in the audio data).") receiving one or more specific textual embeddings from a domain-specific entity retriever based on the plurality of filtered encodings (Wang, Fig. 3, 270; Col. 10, lines 14-21: "In some implementations, the orchestrator component 230 can send the 1-best ASR hypotheses 305a to the ER hypothesis generator 270. The ER hypotheses generator 270, described in further detail below, can generate additional entity mention hypotheses based on the slot text in the 1-best ASR hypothesis 305a. The ER hypothesis generator 270 can return slot text 315b for one or more of its hypotheses to the orchestrator component 230."); providing the plurality of combined embeddings and the one or more specific textual embeddings to a machine-trained model; and receiving a textual output representing speech from the speech input from the machine-trained model (Wang, Fig. 3, 290; Col. 10, lines 36-44: "The orchestrator component 230 can forward the NLU results 210 and the entity data 320 to a skill 290 and/or a skill system 125 for processing. The orchestrator component 230 may receive output data 325 from the skill 290 and/or skill system 125 and cause performance of one or more actions in response to the input audio data 211. In some implementations, the orchestrator component 230 can forward the output data 325 to the TTS component 280 for generating a spoken-word response to the input audio data 211."). However, Wang fails to expressly recite performing blank filtering on a received speech input to generate a plurality of filtered encodings.
Quan teaches performing blank filtering on a received speech input to generate a plurality of filtered encodings (Quan, Page 4, step 130: "step 130, determining the effective frame from the characteristic frame sequence based on the blank probability. In order to reduce the calculation amount of the voice recognition the blank frame can be removed from the characteristic frame sequence, only keeping the effective frame, so as to shorten the characteristic frame sequence for subsequent calculation, so as to improve the calculation efficiency of the voice recognition Specifically, after calculating the blank probability of each characteristic frame, it can determine whether the characteristic frame is a valid frame based on the blank probability.").
Wang and Quan are analogous arts because they each belong to the same field of speech processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the entity resolution system of Wang to incorporate the teachings of Quan to filter blank frames from the audio input. Filtering blank frames reduces the amount of data that must be processed after filtering (Quan, Page 4, step 130). This improves the overall efficiency of the system.
Regarding claim 18, the rejection of claim 17 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. Quan further teaches wherein the plurality of filtered encodings are generated in part using a connectionist temporal classification model (Quan, Page 4, step 120: "step 120, calculating the blank probability of each characteristic frame in the characteristic frame sequence. The Connectionist Temporal Classification (CTC) is a method for avoiding input and output manual alignment."). The same motivation for claim 17 applies equally to claim 18.
Regarding claim 19, the rejection of claim 17 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. Wang further discloses wherein the domain-specific entity retriever is a dual encoder model that comprises keys and values, wherein the keys are acoustic encodings and the values are domain-specific entities (Wang, Col. 12, lines 36-42: "The ER component 266 may query a search index 262 and identify entity data 320 responsive to the audio data 211. The search index 262 may include one or more entity catalogs associated with the intent. The ER component 266 can return the entity data 320 including one or more entities matching the slot text 315.").
Claim(s) 5 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Quan, as applied to claims 1-4, 6-7, 12-15, and 17-19 above, and further in view of Kurata et al. (US Pat. Pub. No. 2022/0208179 A1 hereinafter Kurata).
Regarding claim 5, the rejection of claim 4 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. However, Wang, in view of Quan, fails to expressly recite wherein a text input portion of the connectionist temporal classification model is unused during training.
Kurata teaches wherein a text input portion of the connectionist temporal classification model is unused during training (Kurata, [0044]: "In various embodiments, a Connectionist Temporal Classification (CTC) model is trained with acoustic features, x, represented as vectors as input and phonemes as output to obtain a phoneme acoustic model. The neural network trained with this CTC modeling can be used to initialize the encoder network 130 of RNN-T 100."; [0005]: "The computer implemented method includes synthesizing first domain audio data from first domain text data, and feeding the synthesized first domain audio data into a trained encoder of the recurrent neural network transducer (RNN-T) having an initial condition, wherein the encoder is updated using the synthesized first domain audio data and the first domain text data.").
Wang, Quan, and Kurata are analogous arts because they each belong to the same field of speech processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the entity resolution system of Wang, as modified by the voice coding method of Quan, to incorporate the teachings of Kurata to not use a text input when training the connectionist temporal classification model. This allows a text input to be used separately from the connectionist temporal classification model (Kurata, [0005]). As such, the system is able to receive data even if comparable data was not used for training a portion of the system.
Claim(s) 8 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Quan, as applied to claims 1-4, 6-7, 12-15, and 17-19 above, and further in view of Hu, Fengshuo (US Pat. Pub. No. 2020/0251097 A1 hereinafter Hu).
Regarding claim 8, the rejection of claim 6 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. However, Wang, in view of Quan, fails to expressly recite wherein the plurality of filtered embeddings are provided to the domain-specific entity retriever as the acoustic encodings.
Hu teaches wherein the plurality of filtered embeddings are provided to the domain-specific entity retriever as the acoustic encodings (Hu, [0006]: "According to an aspect of the present disclosure, there is provided a named entity recognition method, including: acquiring a voice signal; extracting a voice feature vector in the voice signal; extracting, based on a literalness result after voice recognition is performed on the voice signal, a literalness feature vector in the literalness result; splicing the voice feature vector and the literalness feature vector to obtain a composite feature vector of each word in the voice signal; processing the composite feature vector of each word in the voice signal through a deep learning model to obtain a named entity recognition result.").
Wang, Quan, and Hu are analogous arts because they each belong to the same field of speech processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the entity resolution system of Wang, as modified by the voice coding method of Quan, to incorporate the teachings of Hu to provide the filtered embeddings to the domain-specific entity retriever as acoustic embeddings. This helps improve recognition precision and accuracy in special scenarios (Hu, [0005]). As such, the system can effectively recognize entities in various specific domains.
Claim(s) 9-10, 16, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Quan, as applied to claims 1-4, 6-7, 12-15, and 17-19 above, and further in view of Dong et al. (US Pat. No. 11,947,912 B1 hereinafter Dong).
Regarding claim 9, the rejection of claim 6 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. However, Wang, in view of Quan, fails to expressly recite wherein the keys and the values are encoded separately and a cosine distance between an encoded key and its respective encoded value is determined to measure a similarity between the encoded key and its respective encoded value.
Dong teaches wherein the keys and the values are encoded separately and a cosine distance between an encoded key and its respective encoded value is determined to measure a similarity between the encoded key and its respective encoded value (Dong, Col. 9, lines 24-30: "For example, the top 5 (or any other number) entity embeddings that are the closest to the encoded representation 314 (e.g., as determined using cosine similarity, Euclidean distance, etc.) may be selected from among the top-k entities returned from among the different memory layers 316. The output may be the entity embedding 318.").
Wang, Quan, and Dong are analogous arts because they each belong to the same field of speech processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the entity resolution system of Wang, as modified by the voice coding method of Quan, to incorporate the teachings of Dong to use a cosine distance between encoded keys and values to identify entities. Identifying entities in this way ensures that the system identifies the most likely entity from amongst numerous possible entities (Dong, Col. 9, lines 24-30). This improves the effectiveness of the entity recognition portion of the system.
Regarding claim 10, the rejection of claim 9 is incorporated. Wang, in view of Quan and Dong, discloses all of the elements of the current invention as stated above. Dong further teaches wherein the one or more specific textual embeddings are determined based on at least one cosine distance determined between a first encoded key and a first respective encoded value (Dong, Col. 9, lines 24-30: "For example, the top 5 (or any other number) entity embeddings that are the closest to the encoded representation 314 (e.g., as determined using cosine similarity, Euclidean distance, etc.) may be selected from among the top-k entities returned from among the different memory layers 316. The output may be the entity embedding 318."). The same motivation for claim 9 applies equally to claim 10.
Regarding claim 16, the rejection of claim 15 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. However, Wang, in view of Quan, fails to expressly recite wherein the keys and the values are encoded separately and a cosine distance between an encoded key and its respective encoded value is determined to measure a similarity between the encoded key and its respective encoded value.
Dong teaches wherein the keys and the values are encoded separately and a cosine distance between an encoded key and its respective encoded value is determined to measure a similarity between the encoded key and its respective encoded value (Dong, Col. 9, lines 24-30: "For example, the top 5 (or any other number) entity embeddings that are the closest to the encoded representation 314 (e.g., as determined using cosine similarity, Euclidean distance, etc.) may be selected from among the top-k entities returned from among the different memory layers 316. The output may be the entity embedding 318.").
Wang, Quan, and Dong are analogous arts because they each belong to the same field of speech processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the entity resolution system of Wang, as modified by the voice coding method of Quan, to incorporate the teachings of Dong to use a cosine distance between encoded keys and values to identify entities. Identifying entities in this way ensures that the system identifies the most likely entity from amongst numerous possible entities (Dong, Col. 9, lines 24-30). This improves the effectiveness of the entity recognition portion of the system.
Regarding claim 20, the rejection of claim 19 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. However, Wang, in view of Quan, fails to expressly recite wherein the keys and the values are encoded separately and a cosine distance between an encoded key and its respective encoded value is determined to measure a similarity between the encoded key and its respective encoded value.
Dong teaches wherein the keys and the values are encoded separately and a cosine distance between an encoded key and its respective encoded value is determined to measure a similarity between the encoded key and its respective encoded value (Dong, Col. 9, lines 24-30: "For example, the top 5 (or any other number) entity embeddings that are the closest to the encoded representation 314 (e.g., as determined using cosine similarity, Euclidean distance, etc.) may be selected from among the top-k entities returned from among the different memory layers 316. The output may be the entity embedding 318.").
Wang, Quan, and Dong are analogous arts because they each belong to the same field of speech processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the entity resolution system of Wang, as modified by the voice coding method of Quan, to incorporate the teachings of Dong to use a cosine distance between encoded keys and values to identify entities. Identifying entities in this way ensures that the system identifies the most likely entity from amongst numerous possible entities (Dong, Col. 9, lines 24-30). This improves the effectiveness of the entity recognition portion of the system.
Claim(s) 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Quan, as applied to claims 1-4, 6-7, 12-15, and 17-19 above, and further in view of Jia et al. (US Pat. Pub No. 2021/0217404 A1 hereinafter Jia).
Regarding claim 11, the rejection of claim 1 is incorporated. Wang, in view of Quan, discloses all of the elements of the current invention as stated above. However, Wang, in view of Quan, fails to expressly recite wherein providing the plurality of combined embeddings and the one or more specific textual embeddings to the machine-trained model comprises prepending the one or more specific textual embeddings to one or more combined embeddings of the plurality of combined embeddings before the machine-learning model processes the plurality of combined embeddings and the one or more specific textual embeddings.
Jia teaches wherein providing the plurality of combined embeddings and the one or more specific textual embeddings to the machine-trained model comprises prepending the one or more specific textual embeddings to one or more combined embeddings of the plurality of combined embeddings before the machine-learning model processes the plurality of combined embeddings and the one or more specific textual embeddings (Jia, [0040]: "The sequence-to-sequence attention neural network may model multiple particular speakers by, for each audio example x in a training dataset, concatenating a d-dimensional embedding vector associated with the true speaker with the output of the encoder neural network at each time step before the output is provide to the attention neural network.").
Wang, Quan, and Jia are analogous arts because they each belong to the same field of speech processing. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the entity resolution system of Wang, as modified by the voice coding method of Quan, to incorporate the teachings of Jia to concatenate the embeddings before they are input into a trained model. This allows the system to combine all relevant information to serve as input for the trained model (Jia, [0040]), allowing the trained model to function effectively.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Goel et al. (US Pat. Pub. No. 2018/0308487 A1) discloses a dialogue system incorporating a unique speech to text conversion method for meaningful dialogue response.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TYLER J BECKER whose telephone number is (703)756-1271. The examiner can normally be reached M-Th, 7:15am-5:45pm PT.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Daniel Washburn can be reached at (571) 272-5551. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/TYLER BECKER/ Examiner, Art Unit 2657
/DANIEL C WASHBURN/ Supervisory Patent Examiner, Art Unit 2657
Prosecution Insights