DETAILED ACTION
This communication is in response to the Application filed on 31 July 2023. Claims 31-45 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 .
Priority
Applicant claims the benefit of US Provisional Application No. 63/145,376, filed 03 February 2021. Claims 31-45 have been afforded the benefit of this filing date.
Information Disclosure Statement
The IDS dated 05 February 2025 has been considered and placed in the application file.
The IDS dated 06 February 2024 has been considered and placed in the application file.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 31-45 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. All of the claims are method claims (31-45), apparatus/machine claims or manufacture claim under (Step 1), but under Step 2A all of these claims recite abstract ideas and specifically mental processes. These mental processes are more particularly recited in claim 31 as:
selecting one or more acoustic or natural language processing (NLP) models…
detecting a result estimating a severity of risk of having the behavioral or mental health condition by processing the input signal with a fused model or joint model derived from the one or more acoustic or NLP models…
Under Step 2A Prong One, claim 31 is directed to an abstract idea and specifically a mental process. As detailed above, the steps of selecting and detecting may be practically performed in the human mind with the use of a physical aid such as a pen and paper. For example, a human could receive input speech, translate the speech to their native language using a dictionary in a case where they do not fluently speak the language associated with the input speech, and based on the translated speech, determine whether the speech indicates mental health risk.
Under Step 2A Prong Two, this judicial exception is not integrated into a practical application because claims 31-45 do not recite additional elements that integrate the exception into a practical application. In particular, claim 31 recites the additional elements of acoustic or NLP models (¶ [0004]), multilingual models (¶ [0078]), and fused or joint models (¶ [0082]). These additional elements are recited at a high level of generality and merely equate to “apply it” or otherwise merely uses a generic computer as a tool to perform an abstract which are not indicative of integration into a practical application as per MPEP 2106.05(f). Further, claim 31 recites the additional elements of “receiving…” which amounts to insignificant extra-solution activities which are not indicative of integration into a practical application as per MPEP 2106.05(g). Accordingly, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
Under Step 2B, the claims do not recite additional elements that are sufficient to amount to significantly more than the judicial exception. The additional limitations in the claims noted above are directed towards insignificant extra-solution activities. The claims are not patent eligible.
With respect to claims 32 and 34, the claim relates to the further specifying particular aspects of the input speech. As described above, the additional elements of “receiving…” amounts to insignificant extra-solution activities which are not indicative of integration into a practical application as per MPEP 2106.05(g). Merely specifying additional elements of the received data does not detract from the abstract idea of the claim as recited above. No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 33 and 35, the claim relates to selecting certain models based on detecting a language of the input speech. This relates to a human detecting the language of the input speech and choosing the right translation dictionary for that speech. The additional element of a “NLP or acoustic model” is recited at a high level of generality (¶ [0004]) and merely equates to “apply it” or otherwise merely uses a generic computer as a tool to perform an abstract which are not indicative of integration into a practical application as per MPEP 2106.05(f). No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 36, the claim relates to applying weights to the output of acoustic or NLP models. This relates to a human weighing different methodologies that diagnose mental health conditions with different weights, such that some methodologies are given more weight than others. The additional element of a “NLP or acoustic model” is recited at a high level of generality (¶ [0004]) and merely equates to “apply it” or otherwise merely uses a generic computer as a tool to perform an abstract which are not indicative of integration into a practical application as per MPEP 2106.05(f). No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 37, the claim relates to using a machine learning model to combine weighted outputs of acoustic or NLP models. This relates to a human following an SVM classifier procedure to combine the weighted scores derived from the various methodologies. The additional elements of a “NLP or acoustic model” and a “machine learning model” are recited at a high level of generality (¶ [0004] and [0077], respectively) and merely equates to “apply it” or otherwise merely uses a generic computer as a tool to perform an abstract which are not indicative of integration into a practical application as per MPEP 2106.05(f). No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 38, the claim relates to selecting a model based on prediction latency, context of the input signal, or demographic information about the subject. This relates to a human choosing between methodologies to analyze a patient’s mental health based on context about the patient such as their body language or mental health history. The additional element of a “NLP or acoustic model” is recited at a high level of generality (¶ [0004]) and merely equates to “apply it” or otherwise merely uses a generic computer as a tool to perform an abstract which are not indicative of integration into a practical application as per MPEP 2106.05(f). No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 39, the claim relates to one or more NLP models including language-specific models, model-translation-based models, input signal-translation-based models, or multilingual models. The additional elements of a “NLP models” and “language-specific models” are recited at a high level of generality (¶ [0004] and [0078], respectively) and merely equates to “apply it” or otherwise merely uses a generic computer as a tool to perform an abstract which are not indicative of integration into a practical application as per MPEP 2106.05(f). No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 40, the claim relates to selecting a model based on confidence measures computed from certainty measures with respect to a language of the input speech. This relates to a human choosing between methodologies to analyze a patient’s mental health based on an estimated confidence level regarding the human’s familiarity with the input speech’s language. The additional element of a “NLP or acoustic model” is recited at a high level of generality (¶ [0004]) and merely equates to “apply it” or otherwise merely uses a generic computer as a tool to perform an abstract which are not indicative of integration into a practical application as per MPEP 2106.05(f). No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 41, the claim relates to iteratively determining remedial actions based on the result of the mental health risk detection result until a reliability threshold is complete. This relates to a human repeatedly asking a patient questions, determining a result related to the patient’s mental health risk, and iteratively performing this process until the human’s diagnosis reaches a threshold level of confidence regarding the results. No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 42, the claim relates to determining remedial actions to improve collection of the input signal. This relates to a human asking the patient to speak louder such that they can better understand the patient’s responses to questions. No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 43, the claim relates to computing certainty measures based on a metadata of the input signal. This relates to a human taking into account the background or demographic of the patient when determining how familiar they are with the patient’s language. No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 44, the claim relates to computing confidence measures based on a the length of a session, a topic of the input signal, or a quality of the input signal. This relates to a human taking into account the quality of the input speech recording when determining how confident they are in the patient’s mental health diagnosis. No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 45, the claim relates to using lexical or audio characteristics of the input signal to determine a weighted combination of the NLP or acoustic models. This relates to a human using the input speech’s volume to determine whether they should diagnose a patient’s mental health risk using a transcript or the audio itself and appropriately weighting the transcript and audio analyses to reflect their decision. No additional limitations are present. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
For all of the above reasons, taken alone or in combination, claims 31-45 recite a non-statutory mental process.
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 31-36, 38-39, and 45 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by US Patent Publication 20190385711 A1 (Shriberg et al.).
Claim 31
Regarding claim 31, Shriberg et al. disclose a method for detecting a behavioral or mental health condition, the method comprising:
(a) receiving an input signal comprising a plurality of audio or lexical characteristics of speech of a subject (Shriberg et al. ¶ [0155], "The clinician 130 may record one or more speech samples from the patient 120 and upload them to the wide area network 110, with the consent of the patient 120. The speech samples may be analyzed by one or more machine learning algorithms, described elsewhere herein."), wherein at least one of the plurality of audio or lexical characteristics of the speech relates to at least one language (Shriberg et al. ¶ [0401], "Speech recognition is specific to the particular language of the speech. Accordingly, speech recognition 2210 includes language-specific speech recognition 3702, which in turn includes a number of language-specific speech recognition engines 3706A-Z.");
(b) based at least in part on the plurality of audio or lexical characteristics of the input signal, selecting one or more acoustic or natural language processing (NLP) models (Shriberg et al. ¶ [0401], "Speech recognition is specific to the particular language of the speech. Accordingly, speech recognition 2210 includes language-specific speech recognition 3702, which in turn includes a number of language-specific speech recognition engines 3706A-Z." See Figure 37. Selecting any one of the language-specific speech recognition engines 3706A-Z to transcribe speech is considered analogous to selecting one or more acoustic or NLP models), wherein selecting one or more acoustic or NLP models comprises at least one of:
selecting a multilingual model,
selecting a model with a model language that includes the at least one language (Shriberg et al. ¶ [0401], "Speech recognition is specific to the particular language of the speech. Accordingly, speech recognition 2210 includes language-specific speech recognition 3702, which in turn includes a number of language-specific speech recognition engines 3706A-Z." See Figure 37, which illustrates a plurality of language-specific speech recognition engines 3706A-Z. Any one of these engines is considered analogous to a model with a model language that includes the at least one language),
selecting a model with a model language and translating the input signal from the at least one language to the model language (Shriberg et al. ¶ [0402], "Suppose for example that the patient speaks a language that is recognized by any of language-specific speech recognition engines 3706A-Z but is not processed by language models 2214 (FIG. 22). Language-specific speech recognition 3702 (FIG. 37) produces text in the language spoken by the patient, i.e., the patient's language, from the audio signal received from the patient. ... translation engine 3704 translates the text from the patient's language to a language that may be processed by language models 2214, e.g., English."), and
selecting a model with a model language and translating the model from the model language to the at least one language; and
(c) detecting a result estimating a severity of risk of having the behavioral or mental health condition (Shriberg et al. ¶ [0420]-[0423], "The models described herein may produce scores, at various stages of an assessment. ... A mental state may be expressed as a composition of mental sub-states, and a patient's composite mental state may be a weighted average of individual scores from the mental sub-states. For example, a composition score of depression may be a weighted average of individual scores for anger, sadness, self-image, self-worth, stress, loneliness, isolation, and anxiety.") by processing the input signal with a fused model or joint model derived from the one or more acoustic or NLP models (Shriberg et al. ¶ [0152], "a server computer system (health screening or monitoring server 102—FIG. 1A) may apply a health state screening or monitoring test to a human patient using a client device (patient device 112), by engaging the patient in an interactive spoken conversation and applying a composite model, that may combine language, acoustic, metadata, and visual models, to a captured audiovisual signal of the patient engaged in the dialogue. ... the composite model may analyze, in real time, the audiovisual signal of the patient (i) to make the conversation more engaging for the patient and (ii) estimate the patient's health." A composite model that comprises language, acoustic, metadata, and visual models is considered analogous to a fused model or joint model derived from one or more acoustic or NLP models).
Claim 32
Regarding claim 32, the rejection of claim 31 is incorporated.
Shriberg et al. further disclose wherein the input signal further comprises a plurality of lexical characteristics of speech of an agent (Shriberg et al. ¶ [0278], "Assessment test administrator 2202 may further distinguish the patient's voice from the clinician's voice using language models 2214, particularly, semantic pattern models such as semantic pattern modules 4004, to identify which of the two distinguished voices more frequently asks questions. Assessment test administrator 2202 may further distinguish the patient's voice from the clinician's voice using acoustic models 2016, which may identify and segment out the clinician's voice from an acoustic analysis of the clinician's voice performed prior to the clinical encounter." A clinician is considered analogous to an agent. Using language or acoustic models to distinguish clinician speech from patient speech implies that the clinician's speech must have a plurality of lexical characteristics to distinguish).
Claim 33
Regarding claim 33, the rejection of claim 31 is incorporated.
Shriberg et al. further disclose detecting the at least one language (Shriberg et al. ¶ [0401], "Speech recognition is specific to the particular language of the speech. Accordingly, speech recognition 2210 includes language-specific speech recognition 3702, which in turn includes a number of language-specific speech recognition engines 3706A-Z."); and
based on the at least one language:
selecting a model with a model language that includes the at least one language when the at least one language is from a database of supported languages; or
selecting a model with a model language that does not include the at least one language when the at least one language is not from a database of supported languages (Shriberg et al. ¶ [0402], "Suppose for example that the patient speaks a language that is recognized by any of language-specific speech recognition engines 3706A-Z but is not processed by language models 2214 (FIG. 22). Language-specific speech recognition 3702 (FIG. 37) produces text in the language spoken by the patient, i.e., the patient's language, from the audio signal received from the patient. ... translation engine 3704 translates the text from the patient's language to a language that may be processed by language models 2214, e.g., English.").
Claim 34
Regarding claim 34, the rejection of claim 31 is incorporated.
Shriberg et al. further disclose wherein the input signal comprises text or audio (Shriberg et al. ¶ [0155], "The clinician 130 may record one or more speech samples from the patient 120 and upload them to the wide area network 110, with the consent of the patient 120.").
Claim 35
Regarding claim 35, the rejection of claim 31 is incorporated.
Shriberg et al. further disclose wherein selecting one or more acoustic or NLP models comprises selecting a model with a model language or dialect that is similar to the at least one language (Shriberg et al. ¶ [0401], "Speech recognition is specific to the particular language of the speech. Accordingly, speech recognition 2210 includes language-specific speech recognition 3702, which in turn includes a number of language-specific speech recognition engines 3706A-Z." See Figure 37. Selecting any one of the language-specific speech recognition engines 3706A-Z to transcribe speech is considered analogous to selecting one or more acoustic or NLP models with a model language or dialect that is similar to the user's language).
Claim 36
Regarding claim 36, the rejection of claim 31 is incorporated.
Shriberg et al. further disclose wherein processing the input signal with a single or fused model derived from the one or more acoustic or NLP models comprises
(i) applying one or more weights to outputs of the one or more acoustic or NLP models (Shriberg et al. ¶ [0308], "Model weighting may be done using static weights, such as weighting the output of the NLP model 2015 more than either the acoustic model 2016 or video model 2017 outputs."); and
(ii) combining the one or more weighted outputs (Shriberg et al. ¶ [0308], "model weight and fusion engine 2019 combines the model outputs into a single consolidated classification for the health state of each client 260a-n.").
Claim 38
Regarding claim 38, the rejection of claim 31 is incorporated.
Shriberg et al. further disclose wherein the selecting the one or more acoustic or NLP models is also based at least in part on a latency of a prediction, a context of the input signal, or demographic information about the subject (Shriberg et al. ¶ [0313], "In some embodiments, if the metadata for an individual provides insight into that person's age, gender, ethnicity, educational background, accent/region they grew up in, etc. this information may be utilized to select the most appropriate model to use in future interactions with this given patient, and may be likewise used to train models that apply to individuals that share similar attributes.") or the joint model or fused model derived from the at least one or more acoustic or NLP models further comprises at least one demographic or metadata-based model (Shriberg et al. ¶ [0305]-[0306], "In FIG. 20B the runtime model server(s) 2010 is provided in greater detail. ... FIG. 20B also includes a metadata model 2018. The metadata model may analyze patient data, such as demographic data, medical history data, and patient-provided data." See Figure 20B. Modelers 2020 is considered analogous to a joint model or fused model derived from at least one or more acoustic or NLP models).
Claim 39
Regarding claim 39, the rejection of claim 31 is incorporated.
Shriberg et al. further disclose wherein the one or more NLP models include one or more of language-specific models (Shriberg et al. ¶ [0401], "Modeling system logic 5320 includes speech recognition 2210 (FIG. 22), which is shown in greater detail in (FIG. 37). Speech recognition is specific to the particular language of the speech. Accordingly, speech recognition 2210 includes language-specific speech recognition 3702, which in turn includes a number of language-specific speech recognition engines 3706A-Z." See Figure 37, which illustrates a plurality of language-specific models), model-translation-based models, input signal-translation-based models (Shriberg et al. ¶ [0402], "Speech recognition 2210 also includes a translation engine 3704." See Figure 37, which illustrates an input signal-translation-based model (e.g. translation engine 3704)), or multilingual models.
Claim 45
Regarding claim 45, the rejection of claim 31 is incorporated.
Shriberg et al. further disclose wherein selecting one or more acoustic or NLP models comprises using the plurality of audio or lexical characteristics of the input signal to determine a weighted combination of the one or more acoustic or NLP models (Shriberg et al. ¶ [0307]-[0308], "modelers 2020 consume the models, preprocessed audio and visual data, and ASR 2012 output to analyze the clients' 260a-n responses for the health state in question. ... The output for each of these modelers 2020 is provided, individually, to a calibration, confidence, and desired descriptors module 2092. ... The output of desired description module 2092 is provided to model weight and fusion engine 2019. This model weight and fusion engine 2019 combines the model outputs into a single consolidated classification for the health state of each client 260a-n. Model weighting may be done using static weights, such as weighting the output of the NLP model 2015 more than either the acoustic model 2016 or video model 2017 outputs.").
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.
Claim 37 is rejected under 35 U.S.C. 103 as obvious over Shriberg et al. as applied to claim 36 above, and further in view of "Ensemble methods in machine learning" (Dietterich).
Claim 37
Regarding claim 37, the rejection of claim 36 is incorporated.
Shriberg et al. do not explicitly disclose all of combining model outputs using a machine learning model.
However, Dietterich discloses wherein combining one or more weighted [acoustic or NLP] models comprises using a machine learning model to combine the outputs of one or more [acoustic or NLP] models (Dietterich pg. 6, Section 2.2, Paragraph 4-5, "The third method for manipulating the training set is illustrated by the AdaBoost algorithm.... AdaBoost maintains a set of weights over the training examples. In each iteration
l
, the learning algorithm is invoked to minimize the weighted error on the training set, and it returns an hypothesis
h
l
. The weighted error of
h
l
is computed and applied to update the weights on the training examples. ... The final classifier ... is constructed by a weighted vote of the individual classifiers. Each classifier is weighted (by
w
l
) according to its accuracy on the weighted training set that it was trained on." A final classifier is considered analogous to a machine learning model used to combine the outputs of one or more models).
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify Shriberg et al.’s model fusion strategy to incorporate Dietterich’s machine-learning-based ensemble technique.
The suggestion/motivation for doing so would have been that, “Several experimental studies have been performed to compare ensemble methods. ... Table 1 ... shows that AdaBoost often gives the best results” as noted by Dietterich in pg. 9, Section 3, Paragraph 1.
Claims 40-44 are rejected under 35 U.S.C. 103 as obvious over Shriberg et al. as applied to claim 31 above, and further in view of US Patent Publication 20220108073 A1 (Menon et al.).
Claim 40
Regarding claim 40, the rejection of claim 31 is incorporated.
Shriberg et al. further disclose prior to (b), determining one or more certainty measures, wherein a certainty measure of the one or more certainty measures corresponds to a language of the input signal (Shriberg et al. ¶ [0350], "The text quality confidence module 2560 determines a confidence measure for the output of the ASR output post processor 2510." ¶ [0430], "A confidence measure may depend on conditions under which the score was taken. ... Conditions may include ... language fluency of a speaker" ASR transcription, when considering conditions that include language fluency, is considered analogous to a certainty measure corresponding to a language of the input signal), and determining at least one confidence measure based on at least one of the one or more certainty measures (Shriberg et al. ¶ [0283]-[0284], "ASR logic 1804 makes [the confidence] information available to runtime models 1802. Runtime models 1802 include descriptive model and analytics 1812, natural language processing (NLP) model 1806, acoustic model 1808, and visual model 1810." ¶ [0311], "The model trainers 2039 consume the ... ASR output data in a NLP trainer 2033, an acoustic trainer 2034, a video trainer 2035, and a metadata trainer 2036. The trained models are provided, individually, to a calibration, confidence, and desired descriptors module 2097. This module calibrates the outputs in order to produce scaled scores, as well as provides confidence measures for the scores."), wherein a confidence measure is associated with a performance, with respect to the language of the at least one certainty measure (Shriberg et al. ¶ [0430], "A confidence measure may depend on conditions under which the score was taken. ... Conditions may include ... language fluency of a speaker"), of an acoustic model or NLP model of a plurality of acoustic or NLP models for detecting the severity of risk of having the behavioral or mental health condition (Shriberg et al. ¶ [0430], "The models described herein may use confidence measures. A confidence measure may be a measure of how effective the score produced by the machine learning algorithm may be in order of accurately predicting a mental state, such as depression.")….
Shriberg et al. do not explicitly disclose all of selecting models based on a confidence measure.
However, Menon et al. discloses [in (b),] selecting one or more acoustic or NLP models [is] based at least in part on [the] at least one confidence measure (Menon et al. ¶ [0033], "entity classifier also designates a confidence score indicating the likelihood that the classifier's labeling is correct. The predicted entities block 116 may select from the classifier-labeled entities set entities that are associated with a high confidence score (e.g., greater than 50%, 60%, 75%, 95%, etc.), for another iteration of the training process, as indicated by the path 118.").
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify Shriberg et al.’s language-based model selection to include Menon et al.’s confidence measures because such a modification is the result of combining prior art elements according to known methods to yield predictable results. More specifically, Shriberg et al.’s language-based model selection as modified by Menon et al.’s confidence-based model selection can yield a predictable result of improving accuracy since model selection could more informatively choose a model based on both confidence and language identification. Thus, a person of ordinary skill would have appreciated including in Shriberg et al.’s language-based model selection the ability to do Menon et al.’s confidence measures since the claimed invention is merely a combination of old elements, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Claim 41
Regarding claim 41, the rejection of claim 31 is incorporated.
Shriberg et al. further disclose (d) [based on a reliability measure of the result,] iteratively or dynamically selecting one or more remedial actions (Shriberg et al. ¶ [0466]-[0467], "The following are examples of alert rules that may be implemented by interactive screening or monitoring server logic 502. ... In an audio diary in which a patient is detected to be at least moderately depressed, interactive screening or monitoring server logic 502 may record the analysis in the patient's clinical records such that the patient's behavioral health care provider may discuss the diary entry when the patient is next seen." Recording patient analysis to a patient's clinical records is considered analogous to a remedial action) and repeating (b) and (c) (Shriberg et al. ¶ [0465]-[0468], "In step 4710, generalized dialogue flow logic 602 sends the captured speech, along with any context determined in prior yet contemporary performances of step 4708 or step 4710, to runtime model server logic 504 for analysis and evaluation. ... Processing according to the loop of steps 4702-4714 (FIG. 47) continues until stopped by ... monitoring server logic 502" ¶ [0284], "Runtime models 1802 include descriptive model and analytics 1812, natural language processing (NLP) model 1806, acoustic model 1808, and visual model 1810." See Figure 47; iteratively analyzing user speech with runtime models is considered analogous to repeating steps (b) and (c)) [until a reliability threshold is achieved].
Shriberg et al. do not explicitly disclose all of iterative processing based on a reliability measure meeting a reliability threshold.
However, Menon et al. discloses [in (b),] based on a reliability measure of the result, iteratively or dynamically [selecting one or more remedial actions and repeating (b) and (c)] processing until a reliability threshold is achieved (Menon et al. ¶ [0054], "Steps 204 through 212 are repeated until the termination condition is reached.... When the termination condition is reached, a final entity-prediction ML model is obtained along with a labeled corpus, in step 214. ... Termination condition may also include determining whether the number of classifier-labeled entities having a high confidence score is at least equal to a specified labeled-entity-increment threshold. If one or a specified number of iterations do not produce classifier-labeled entities having a high confidence score at least equal to the labeled-entity-increment threshold, the iterations may be terminated." See Figure 2, which illustrates iteratively processing (e.g. classifying (208) and predicting (210)) based on a reliability measure (e.g. "confidence score") until a reliability threshold is achieved (e.g. "labeled-entity-increment threshold")).
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify Shriberg et al.’s iterative technique to include Menon et al.’s reliability measure and threshold because such a modification is the result of combining prior art elements according to known methods to yield predictable results. More specifically, Shriberg et al.’s iterative technique as modified by Menon et al.’s reliability measure and threshold can yield a predictable result of improving a diagnosis’ credibility since the a reliability threshold could set a bare minimum level of confidence associated with a mental health condition diagnosis. Thus, a person of ordinary skill would have appreciated including in Shriberg et al.’s iterative technique the ability to do Menon et al.’s reliability measure and threshold since the claimed invention is merely a combination of old elements, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Claim 42
Regarding claim 42, the rejection of claim 41 is incorporated.
Shriberg et al. further disclose (e) based at least in part on the result, dynamically determining one or more remedial actions to the subject or to an agent to improve collection of the input signal (Shriberg et al. ¶ [0240]-[0242], "Loop step 1504 and next step 1510 define a loop in which assessment test administrator 2202 processes the audiovisual signal according to steps 1506-1508 until assessment test administrator 2202 determines that the audiovisual signal is of high quality or at least of adequate quality to provide accurate assessment. ... If the audiovisual signal received from patient device 312 is determined by assessment test administrator 2202 to be of inadequate quality, the message asks the patient to adjust her environment to improve the signal quality. For example, if the audio portion of the signal is poor, the message may be “I'm having trouble hearing you. may you move the microphone closer to you or find a quieter place?”" An administrator recommending a remedial action (e.g. moving the microphone closer) based on the action possibly improving assessment results is considered analogous to determining remedial actions based at least in part on the result that estimates a patient's mental health).
Claim 43
Regarding claim 43, the rejection of claim 40 is incorporated.
Shriberg et al. further disclose wherein the certainty measure is generated at least in part using metadata of the input signal (Shriberg et al. ¶ [0350], "The text quality confidence module 2560 determines a confidence measure for the output of the ASR output post processor 2510. The confidence measure may be determined based on text metadata (demographic information about the patient, environmental conditions, method of recording, etc.)").
Claim 44
Regarding claim 44, the rejection of claim 40 is incorporated.
Shriberg et al. further disclose wherein the at least one confidence measure is based at least in part on a length of a session in which the input signal is recorded, a topic of the input signal, or a quality of the input signal (Shriberg et al. ¶ [0430], "A confidence measure may be a measure of how effective the score produced by the machine learning algorithm may be in order of accurately predicting a mental state, such as depression. A confidence measure may depend on conditions under which the score was taken. ... Under conditions where the quality of the signal or speech makes it more difficult for the speech to be analyzed, the confidence measure may have a smaller value.").
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JACOB B VOGT whose telephone number is (571)272-7028. The examiner can normally be reached Monday - Friday 9:30am - 7pm EST.
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, Paras D Shah can be reached at (571)270-1650. 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.
/JACOB B VOGT/Examiner, Art Unit 2653
/Paras D Shah/ Supervisory Patent Examiner, Art Unit 2653
09/28/2026
