DETAILED ACTION
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 .
Claims 1-20 are pending and have been examined. Claims 1, 10, and 18 are independent.
This Application was published as U.S. 2025/0037702 A1.
Apparent priority: July 27, 2023.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on July 27, 2023 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Drawings
The drawings are objected to because:
In Figure 7, block 710 repeats the text of block 708, "Generate a training data set using the at least one property of each one of the plurality of audio inputs and the plurality of labels." The Specification describes process 700 as proceeding from block 708 to block 710, at which the neural network is trained using the training data set.
In Figure 1, element 142 is labeled "HOT PHYSICAL MACHINE SET 142." The Specification describes element 142 as host physical machine set 142.
To overcome the objection, amend the legend of block 710 of Figure 7 and the label of element 142 of Figure 1 to agree with the Specification.
Corrected drawing sheets in compliance with 37 CFR 1.121(d) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. The figure or figure number of an amended drawing should not be labeled as "amended." If a drawing figure is to be canceled, the appropriate figure must be removed from the replacement sheet, and where necessary, the remaining figures must be renumbered and appropriate changes made to the brief description of the several views of the drawings for consistency. Additional replacement sheets may be necessary to show the renumbering of the remaining figures. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either "Replacement Sheet" or "New Sheet" pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
Claim Objections
Claims 19 and 20 are objected to because of the following informalities: in each claim, "claim 9" should apparently read "claim 18." Claim 9 is a method claim, and claims 19 and 20 are computer program product claims that correspond to claims 6 and 7 and further limit the computer program product of claim 18. For examination, claims 19 and 20 are treated as depending from claim 18.
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 7-9 and 17 are 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 7 recites "receiving, by a processor, a plurality of training audio inputs," and claim 1, from which claim 7 depends, recites "receiving, by a processor, an audio input." It is unclear whether the processor of claim 7 is the processor of claim 1 or a different processor, and therefore which processor "the processor" refers to in the remaining steps of claim 7 and in claim 8. Claim 9 depends from claim 8 and is rejected for the same reason. For examination, the processor of claim 7 is interpreted as the processor of claim 1.
Claim 17 recites "tune a plurality of weights of the neural network," and claim 10, from which claim 17 depends through claim 16, recites "a memory configured to store a plurality of weights of a neural network." It is unclear whether the plurality of weights of claim 17 is the plurality of weights stored in the memory of claim 10 or a different plurality of weights. For examination, the plurality of weights of claim 17 is interpreted as the plurality of weights of claim 10.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless -
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-4 and 18 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Aher (U.S. 11,443,731), hereinafter Aher.
Regarding claim 1, Aher discloses:
1. A computer-implemented method comprising:
"a voice application may perform process 600, implemented on any suitable hardware" (Aher, col. 19, ll. 29-31.)
Process 600, performed by the voice application on hardware, is the computer-implemented method of the claim.
receiving, by a processor, an audio input from a user;
"At step 602, the voice application receives a voice input." (Aher, col. 19, l. 36.)
"A user may assert voice input 101 which includes speech" (Aher, col. 4, l. 17.)
"control circuitry 404, which includes processing circuitry 406 and storage 408" (Aher, col. 13, ll. 61-62.)
The voice input the user asserts is the audio input from a user, and processing circuitry 406 of the control circuitry that runs the voice application is the processor.
running, by the processor, a neural network using at least one property of the audio input to infer at least one audio parameter;
"At step 604, the voice application calculates at least one prosodic metric of the voice input." (Aher, col. 21, ll. 9-10.)
"the system may train a model based on a training set of data, and then apply the model to voice inputs to generate responses having desired prosody" (Aher, col. 2, ll. 50-52.)
"uses machine learning including Convolutional Neural Networks or Recurrent Neural Network or Long Short-Term Memory" (Aher, col. 4, ll. 64-66.)
"The input and outputs of the model may be multi-labelled to generate a predicted pitch and prosody that will engage the user." (Aher, col. 5, ll. 2-4.)
The prosodic metric calculated from the voice input is the at least one property of the audio input, the model, a convolutional neural network, recurrent neural network, or long short-term memory network, is the neural network, and the predicted pitch and prosody the model outputs is the at least one audio parameter it infers.
generating, by the processor, a simulated voice using the at least one audio parameter; and
"At step 608, the voice application generates a synthesized speech response having prosodic characteristics based on the response and based on the at least one prosodic metric." (Aher, col. 21, ll. 48-50.)
"the voice application determines predicted prosodic characteristics of the response using a model, and modifies the predicted prosodic characteristics to generate the prosodic characteristics of the synthesized speech response" (Aher, col. 21, ll. 55-59.)
The synthesized speech response carrying the predicted prosodic characteristics is the simulated voice generated using the at least one audio parameter.
responding, by the processor, to the audio input using the simulated voice.
"At step 610, the voice application causes the synthesized speech response to be outputted." (Aher, col. 22, ll. 5-6.)
"made to sound more natural or otherwise match the prosody of the voice input by applying the determined prosodic characteristics to the response" (Aher, col. 1, l. 66 to col. 2, l. 1.)
Outputting the synthesized speech response to the voice input is responding to the audio input using the simulated voice.
Regarding claim 2, Aher discloses:
2. The computer-implemented method of claim 1, wherein the at least one property of the audio input comprises at least one of a pitch, a speed, a number of pauses, and a frequency of pauses of the audio input.
"The at least one prosodic metric may include pitch, note, duration, prominence, timbre, rate, rhythm" (Aher, col. 21, ll. 10-12.)
Pitch is the recited pitch, and rate is the recited speed, satisfying the at-least-one-of form.
Regarding claim 3, Aher discloses:
3. The computer-implemented method of claim 1, further comprising identifying the at least one property of the audio input using a voice analysis program to identify the at least one property.
"Prosodic engine 223 is configured to determine one or more prosodic metrics associated with a word, group of words, or a voice input." (Aher, col. 8, ll. 39-41.)
"Prosodic engine 223 may include, for example, temporal and spectral analyzers for extracting information about an audio file." (Aher, col. 8, ll. 41-43.)
Prosodic engine 223, with its temporal and spectral analyzers, is the voice analysis program that identifies the at least one property.
Regarding claim 4, Aher discloses:
4. The computer-implemented method of claim 1, wherein the at least one audio parameter of the simulated voice comprises at least one of a pitch, a speed, a voice style, a number of pauses, and a frequency of pauses of the simulated voice.
"The input and outputs of the model may be multi-labelled to generate a predicted pitch and prosody that will engage the user." (Aher, col. 5, ll. 2-4.)
"to specify voice details (e.g., male/female voice, accent, rate, emphasis, or other details)" (Aher, col. 9, ll. 1-2.)
The predicted pitch is a pitch of the simulated voice, the rate is its speed, and the male or female voice is a voice style, satisfying the at-least-one-of form.
Claim 18 is a computer program product claim with limitations corresponding to the limitations of method Claim 1 and is rejected under similar rationale. Additionally:
Regarding claim 18, Aher discloses:
18. A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable by a device to cause the device to: receive an audio input from a user; run a neural network using at least one property of the audio input to infer at least one audio parameter; generate a simulated voice using the at least one audio parameter; and respond to the audio input using the simulated voice.
"An application and/or any instructions for performing any of the embodiments discussed herein may be encoded on computer-readable media." (Aher, col. 15, ll. 54-57.)
"non-transitory including, but not limited to, volatile and nonvolatile computer memory or storage devices" (Aher, col. 15, ll. 60-62.)
The non-transitory computer-readable media on which the instructions for the embodiments are encoded are the computer readable storage medium having program instructions embodied therewith.
Claim Rejections - 35 USC § 103
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.
Claims 5-7, 10-16, 19, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Aher in view of Austraat (U.S. 12,266,340), hereinafter Austraat.
Regarding claim 5, Aher does not teach the classifier and context aspects of the claim. Austraat teaches:
The context of the claim is read, consistent with the Specification (Figure 4, extraction of context 406: use case, situation), as the use case or situation of the audio input.
5. The computer-implemented method of claim 1, further comprising: inputting the at least one property of the audio input into a classifier; and
"applying the input audio data to at least one trained model (e.g., a NLU system) to categorize one or more communication elements of the plurality of communication elements, where the categorizing includes assigning at least one contextual category" (Austraat, col. 45, ll. 34-38.)
"(e.g., word(s), utterance, speech patterns that include a prosody element such as timbre, tone, cadence, pitch, volume, pronunciation, speaking rate, articulation, fluency, intensity, inflection, resonance, etc.)" (Austraat, col. 45, ll. 41-45.)
The trained model to which the input audio data, including its pitch and speaking rate, is applied is the classifier into which the at least one property of the audio input is input.
obtaining, from the classifier, context data that identifies the context of the audio input.
"a context of the audio input is classified, where the context indicates a purpose (e.g., inquiry, comment, request, etc.) for receiving the audio input" (Austraat, col. 48, ll. 44-46.)
The classified context indicating the purpose of the audio input (inquiry, comment, request) is the context data that identifies the context of the audio input.
Aher and Austraat pertain to generating a synthesized speech response to a user's spoken input. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Aher to input the prosodic metrics of the voice input into a trained classifier that classifies the context of the input, including its purpose, as Austraat does. One of ordinary skill in the art would have been motivated to make this modification in order to respond with a speech pattern suited to the purpose of the input, as Austraat teaches:
"a speech pattern that is predicted to facilitate an optimal outcome (e.g., satisfied customer, inquiry resolution, etc.)" (Austraat, col. 44, ll. 27-29.)
Regarding claim 6, Aher discloses:
6. The computer-implemented method of claim 1, further comprising: converting, by the processor, the audio input into a text input;
"ASR module 110 is configured to sample, condition, and digitize the received audio input and analyze the resulting audio file to generate a text input." (Aher, col. 4, ll. 20-23.)
The digitized audio input converted into a text input by ASR module 110 is the conversion of the audio input into a text input.
generating, by the processor, a text response to the text input;
"Response engine 222 is configured to identify a text input and a text response to the text input." (Aher, col. 7, ll. 30-31.)
extracting, by the processor, context data from the text input and the text response;
"The voice application may determine the emotion metric based on the voice input (e.g., keywords, or prosodic metrics of the voice input), based on a response to the voice input (e.g., keywords, or prosodic metrics of the response)" (Aher, col. 22, ll. 48-51.)
Determining the emotion metric from the keywords of the voice input and of the response is extracting data from the text input and the text response.
inputting, by the processor, the context data and the at least one property of the audio input into the neural network to infer the at least one audio parameter; and
"module 350 is configured to tabulate, at least, word index, time interval (e.g., word temporal arrangement in the signal), pitch values, note values, rate values, timber values, emotional metric values" (Aher, col. 11, ll. 30-34.)
"training model 370 may accept as input an emotion metric" (Aher, col. 12, l. 4.)
Training model 370 accepts the pitch and rate values, which are the at least one property of the audio input, as input together with the emotional metric values drawn from the text.
converting, by the processor, the text response into an audio response having the simulated voice.
"speech generator 225 includes a text-to-speech engine configured to identify a text string to be synthesized as a synthesized speech response" (Aher, col. 8, ll. 61-63.)
"At step 608, the voice application generates a synthesized speech response having prosodic characteristics based on the response and based on the at least one prosodic metric." (Aher, col. 21, ll. 48-50.)
The text-to-speech engine synthesizing the response text with the predicted prosody converts the text response into an audio response having the simulated voice.
Aher does not teach the context-data aspects of the claim. Austraat teaches:
The context data is read, consistent with the Specification (Figure 4, extraction of context 406: use case, situation), as data identifying the use case or situation of the audio input.
extracting, by the processor, context data from the text input and the text response;
"applying the input audio data to at least one trained model (e.g., a NLU system) to categorize one or more communication elements of the plurality of communication elements, where the categorizing includes assigning at least one contextual category" (Austraat, col. 45, ll. 34-38.)
"(e.g., word(s), utterance, speech patterns that include a prosody element such as timbre, tone, cadence, pitch, volume, pronunciation, speaking rate, articulation, fluency, intensity, inflection, resonance, etc.)" (Austraat, col. 45, ll. 41-45.)
"a context of the audio input is classified, where the context indicates a purpose (e.g., inquiry, comment, request, etc.) for receiving the audio input" (Austraat, col. 48, ll. 44-46.)
The contextual category that Austraat's trained model assigns to the communication elements of the input, including its words, which indicates the purpose of the input, is context data.
Aher and Austraat pertain to generating a synthesized speech response to a user's spoken input. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Aher to extract, from the text input and the text response from which it draws the emotion metric, the context of the input, its purpose, as Austraat's trained model does from the words of the user's input, and to input that context data with the prosodic metrics into Aher's neural network, as Aher inputs the emotion metric. One of ordinary skill in the art would have been motivated to make this modification in order to respond with a speech pattern suited to the purpose of the input, as Austraat teaches:
"a speech pattern that is predicted to facilitate an optimal outcome (e.g., satisfied customer, inquiry resolution, etc.)" (Austraat, col. 44, ll. 27-29.)
Regarding claim 7, Aher discloses:
7. The computer-implemented method of claim 1, further comprising: receiving, by a processor, a plurality of training audio inputs from a plurality of users;
"At step 702, the application receives a plurality of voice inputs, each associated with at least one respective voice input prosodic metric." (Aher, col. 23, ll. 46-48.)
"audio database 310 may include a plurality of voice inputs and corresponding responses, indexed or categorized by user" (Aher, col. 10, ll. 8-10.)
"prosodic information 170 includes statistical information for a plurality of users" (Aher, col. 5, ll. 37-38.)
The plurality of voice inputs, held by user and summarized for a plurality of users, are the plurality of training audio inputs from a plurality of users.
receiving, by the processor, a plurality of labels from the plurality of users, wherein the plurality of labels indicate simulated voice prosody preferences of the plurality of users;
"audio database 310 may include metadata tags identifying one or more prosodic metrics associated with each voice input and response" (Aher, col. 10, ll. 5-7.)
extracting, by the processor, at least one property from each one of the plurality of training audio inputs;
"Prosodic engine 223 may be configured to extract a set of information per audio utterance" (Aher, col. 9, ll. 12-13.)
"module 350 is configured to tabulate, at least, word index, time interval (e.g., word temporal arrangement in the signal), pitch values, note values, rate values, timber values, emotional metric values" (Aher, col. 11, ll. 30-34.)
generating, by the processor, a training data set using the at least one property of each one of the plurality of training audio inputs and the plurality of labels; and
"module 350 generates an input file, having a predetermined format, for use as input to training model 370" (Aher, col. 11, ll. 28-30.)
"Training model 370 is configured to accept as input data from module 350, extracted from data of audio database 310; perform a computation; and provide output 380. For example, the data provided as input by module 350 (e.g., the 'training data') is extracted for a set of voice inputs and responses." (Aher, col. 11, ll. 36-41.)
"The input and outputs of the model may be multi-labelled to generate a predicted pitch and prosody that will engage the user." (Aher, col. 5, ll. 2-4.)
The training data that module 350 generates from the features extracted for each voice input, with the inputs of the model labelled, is the training data set generated using the at least one property of each one of the plurality of training audio inputs and the plurality of labels. Module 350 generates it for use as input to training model 370, so it is generated before the model is trained on it.
training, by the processor, the neural network using the training data set.
"At step 706, the application trains the model based on the plurality of voice inputs, the plurality of responses, the voice input prosodic metrics, and the response prosodic metrics" (Aher, col. 23, ll. 64-66.)
"Training model 370 executes an algorithm (e.g., 'trains' the data), and generates output labels (e.g., output 380)." (Aher, col. 11, ll. 47-48.)
Training model 370, trained on the training data that module 350 provides, is the neural network trained using the training data set.
Aher does not teach the user-preference aspects of the labels. Austraat teaches:
The labels are read, under their broadest reasonable interpretation, as indications given by the users of how they regard the prosody of a simulated voice (Specification, Figure 3, training labels 308).
receiving, by the processor, a plurality of labels from the plurality of users, wherein the plurality of labels indicate simulated voice prosody preferences of the plurality of users;
"the measurement of the user’s response compares information provided via a verbal response (e.g., based on a scaled net promotor score (NPS)) from the user to obtain the quantifiable quality score" (Austraat, col. 47, ll. 41-44.)
"the computing system continuously modifies the prosodic configuration of the deployed text-to-speech model based on receiving future input audio data, providing one or more future audio outputs, and identifying future reactions of the user" (Austraat, col. 47, ll. 31-37.)
"the profile(s) may be contextually classified based on one or more purposes for user interactions between previous users and the computing system" (Austraat, col. 49, ll. 23-25.)
Each user's own rating of the synthesized speech pattern, the scaled net promoter score used to modify the prosodic configuration of the text-to-speech model, gathered across the interactions of previous users, is a label from the users indicating their simulated voice prosody preferences.
Aher and Austraat pertain to generating a synthesized speech response to a user's spoken input. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Aher to receive the users' ratings of the synthesized prosody, as Austraat does, and to use them as the labels with which module 350 generates Aher's training data from its plurality of users, before training model 370 is trained on it. One of ordinary skill in the art would have been motivated to make this modification in order to train the model toward the speech pattern the users respond to best, as Austraat teaches:
"a speech pattern that is predicted to facilitate an optimal outcome (e.g., satisfied customer, inquiry resolution, etc.)" (Austraat, col. 44, ll. 27-29.)
Claim 10 is a system claim with limitations corresponding to the limitations of Claim 1 and is rejected under similar rationale. Additionally:
Regarding claim 10, Aher discloses:
10. A system comprising: a memory configured to store a plurality of weights of a neural network; a processor configured to: receive an audio input from a user; run the neural network using at least one property of the audio input to infer at least one audio parameter; generate a simulated voice using the at least one audio parameter; and respond to the audio input using the simulated voice.
"control circuitry 404, which includes processing circuitry 406 and storage 408" (Aher, col. 13, ll. 61-62.)
"the application may store the model and the reference information" (Aher, col. 24, ll. 25-26.)
Processing circuitry 406 is the processor, and storage 408, which stores the trained model, is the memory.
Aher does not teach the weight-storage aspects of the claim. Austraat teaches:
a memory configured to store a plurality of weights of a neural network;
"the system may include one or more memory devices and/or storage devices (represented by memory 524 in FIG. 5) for processing use and/or including one or more instructions necessary for operation of the AI program 502. In some embodiments, the AI program 502 may include a deep neural network" (Austraat, col. 34, ll. 41-47.)
"labels W1 and W2 refer to respective assigned weights for the referenced connections. Two hidden nodes 283 and 285 share the same set of weights W1 and W2 when connecting to two local patches." (Austraat, col. 32, ll. 33-37.)
Memory 524 holding the deep neural network of the AI program, whose connections carry assigned weights, is a memory configured to store a plurality of weights of a neural network.
Aher and Austraat pertain to generating a synthesized speech response to a user's spoken input. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to store Aher's trained neural network model as the weights of its connections in memory, as Austraat does. One of ordinary skill in the art would have been motivated to make this modification because the trained network must be held in memory for the voice application to run it, as Austraat teaches of its memory 524, which includes what is "necessary for operation of the AI program 502" (Austraat, col. 34, ll. 44-45.).
Claim 11 is a system claim with limitations corresponding to the limitations of Claim 2 and is rejected under a similar rationale.
Claim 12 is a system claim with limitations corresponding to the limitations of Claim 3 and is rejected under a similar rationale.
Claim 13 is a system claim with limitations corresponding to the limitations of Claim 4 and is rejected under a similar rationale.
Claim 14 is a system claim with limitations corresponding to the limitations of Claim 5 and is rejected under a similar rationale.
Claim 15 is a system claim with limitations corresponding to the limitations of Claim 6 and is rejected under a similar rationale.
Claim 16 is a system claim with limitations corresponding to the limitations of Claim 7 and is rejected under a similar rationale.
Claim 19 is a computer program product claim with limitations corresponding to the limitations of Claim 6 and is rejected under a similar rationale.
Claim 20 is a computer program product claim with limitations corresponding to the limitations of Claim 7 and is rejected under a similar rationale.
Claims 8, 9, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Aher in view of Austraat, and further in view of Zhang (U.S. 11,996,084), hereinafter Zhang.
Regarding claim 8, Aher and Austraat do not teach the recited training until the error converges and the deployment in response to that convergence. Zhang teaches:
The error converging to a predefined value is read, consistent with the Specification (Figure 3, error 330), as the training error of the neural network reaching a preset level.
8. The computer-implemented method of claim 7, further comprising: tuning, by the processor, a plurality of weights of the neural network until an error of the neural network converges to a predefined value; and
"model parameters of the speech synthesis model may be updated iteratively by means of gradient descent until an iteration stop condition is reached" (Zhang, col. 9, ll. 36-39.)
"the value of the loss function meeting a preset requirement" (Zhang, col. 9, ll. 40-41.)
"takes minimizing a difference between the predicted vocoder feature and the annotated vocoder feature as a training objective" (Zhang, col. 9, ll. 32-35.)
The model parameters of Zhang's neural speech synthesis model are the plurality of weights of the neural network, the loss function measuring the difference between the predicted and the annotated features is its error, and updating the parameters until the value of the loss function meets a preset requirement is tuning the weights until the error converges to a predefined value.
in response to the error converging to the predefined value, deploying the neural network.
"The speech synthesis model obtained after the training may be used in the above embodiment of the speech synthesis method" (Zhang, col. 10, ll. 4-6.)
Zhang puts the model obtained after the training, which ends when the stop condition is reached, to use in its speech synthesis method, which is deploying the neural network.
Aher, Austraat, and Zhang pertain to neural network speech synthesis. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to train the neural network of the combination of Aher and Austraat until the value of its loss function meets a preset requirement, and then to put the trained model to use, as Zhang does. Deploying the model upon reaching the stop condition, as the condition for its deployment, is the timing this combination supplies. One of ordinary skill in the art would have been motivated to make this modification in order to obtain a trained model of the quality Zhang describes:
"The above speech synthesis model has low requirements on training data, and generally can achieve commercial stability, expressiveness and fluency through hundreds of sentences." (Zhang, col. 13, ll. 30-33.)
Regarding claim 9, Aher discloses:
9. The computer-implemented method of claim 8, wherein deploying the neural network comprises deploying the neural network to a voicebot application.
"the results of process 700 may be used as reference information" (Aher, col. 23, ll. 36-37.)
"personal voice assistants such as, for example, Alexa (Amazon), Cortana (Microsoft), and Siri (Apple)" (Aher, col. 4, ll. 1-3.)
Aher's voice application, which uses the results of the training process, is provided by personal voice assistants that converse with the user by voice, which are voicebot applications. The rationale for the combination is similar to the one provided for claim 8.
Claim 17 is a system claim with limitations corresponding to the limitations of Claim 8 and is rejected under a similar rationale.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
a. Nicolis et al. (U.S. 11,062,694): speech-quality models classify paralinguistic features of a user's speech, such as whispered or shouted delivery, and the synthesized speech output is configured to correspond to the indicated quality.
b. Shechtman et al. (U.S. 10,418,025): an expressive prosody model trained with a machine learning module on target and derived neutral prosody vector sequences and used within a text-to-speech system.
c. Alders et al. (U.S. 10,521,512): a smart speaker measures the volume, intonation, and speed of a spoken query and formats its response in a corresponding tone, such as a simulated whisper or shout.
d. Le Gallo-Bourdeau et al. (U.S. 11,386,319): training the weights of a neural network until a convergence condition is met and using the trained network.
e. Cygert et al. (U.S. 12,087,270): user ratings of a customized synthetic voice and the preference data of similar users are used to determine the synthetic voice generated for the user.
f. Zhao et al. (U.S. 2021/0225357): intent recognition from recognized text and acoustic feature annotations of speech input, and an emotional text-to-speech model adapted from labelled training examples.
g. Peddinti et al. (U.S. 2022/0284882): selecting between a user's pronunciation and a text-to-speech pronunciation of a word for a synthesized response during a dialog.
h. Aher et al. (U.S. 2021/0319780): generating synthesized speech responses to voice inputs based on the prosodic metrics of the voice inputs.
i. Bakis et al. (U.S. 8,438,032): a tool with which a user tunes the pitch and duration targets and the prosody of synthesized speech, including prosody supplied by a sample recording.
j. Raitio et al. (U.S. 2017/0358301): a digital assistant that determines from a user's speech input that a whispered speech response is to be provided and provides it.
k. Richardson et al. (U.S. 5,699,487): an artificial neural network trained by iterative backpropagation until it converges.
l. Lepsoy et al. (WO 2023/215132): interactive modification of the speaking style of synthesized speech.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to YUVAL H. LEVENTAL whose telephone number is (571) 270-3130. The examiner can normally be reached Monday-Friday, 8:00 AM - 5:00 PM.
If attempts to reach the examiner by telephone are unsuccessful, the examiner's supervisor, PIERRE-LOUIS DESIR, can be reached at (571) 272-7799. 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.
/YUVAL HAIM LEVENTAL/
Examiner, Art Unit 2659
/PIERRE LOUIS DESIR/Supervisory Patent Examiner, Art Unit 2659
Prosecution Insights