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 .
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. Such claim limitation(s) is/are:
a bias identification engine configured to: retrieve… in claim 9.
the bias identification engine is further configured to: determine… in claim 10.
the bias identification engine is further configured to: generate… in claim 11.
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
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 1-20 are rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1, in step 1 of the 101 analysis set forth in MPEP 2106, the claim recites A method for dynamically identifying bias in a dataset,. The claim recites a method. A method is one of the four statutory categories of invention.
In Step 2A, Prong 1 of the 101 analysis set forth in MPEP 2106, the examiner has determined that the following limitations recite a process that, under broadest reasonable interpretation, covers a mental process or mathematical concept but for the recitation of generic computer components:
during a model training session, determining,…and using a Uniform Discretized Integrated Gradient (UDIG) technique, that a data element corresponds to biased data; (i.e., the broadest reasonable interpretation includes a step of observation, evaluation, and judgement and could be performed mentally or with pen and paper like determining a an input is biased, which is either a mental process of observation/evaluation/judgement (MPEP 2106). Additionally, the limitation also recites the use of an integrated technique. Under the broadest reasonable interpretation, the limitation includes a mathematical calculation like calculating an integrated gradient, a mathematical calculation is considered a mathematical concept (MPEP 2106)).
in response to determining that the data element corresponds to biased data, determining…a bias identification event, wherein the bias identification event corresponds to a bias identification event type and the bias identification event type is based on the data element that corresponds to the biased data; (i.e., the broadest reasonable interpretation includes a step of observation, evaluation, and judgement and could be performed mentally or with pen and paper like determining a category of a biased input, which is either a mental process of observation/evaluation/judgement (MPEP 2106)).
determining,…and based on the bias identification event type, a bias mitigation action; (i.e., the broadest reasonable interpretation includes a step of observation, evaluation, and judgement and could be performed mentally or with pen and paper like determining a mitigation action, which is either a mental process of observation/evaluation/judgement (MPEP 2106)).
If the claim limitations, under their broadest reasonable interpretation, covers activities classified under Mental processes: concepts performed in the human mind (including observation, evaluation, judgement, or opinion) (see MPEP 2106.04(a)(2), subsection (III)) or Mathematical concepts: mathematical relationships, mathematical formulas or equations, or mathematical calculations (see MPEP 2106.04(a)(2), subsection (I)). Accordingly, the claim recites an abstract idea.
In Step 2A, Prong 2 of the 101 analysis, set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
the method comprising: receiving, by communications hardware, a fine-tuning request, wherein the fine-tuning request comprises a set of fine-tuning parameters; (i.e., the broadest reasonable interpretation of receiving a data instance is mere data gathering, which is an insignificant extra solution activity (MPEP 2106.05(g))).
retrieving, by a bias identification engine and based on the set of fine-tuning parameters, a machine learning model and a training dataset, wherein the training dataset comprises a plurality of data elements; (i.e., the broadest reasonable interpretation of receiving a data instance is mere data gathering, which is an insignificant extra solution activity (MPEP 2106.05(g))).
…by the bias identification engine… (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
…by bias treatment circuitry… (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
and causing, by the bias treatment circuitry, performance of the bias mitigation action. (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
Since the claim does not contain any other additional elements, that amount to integration into a practical application, the claim is directed to an abstract idea.
In Step 2B of the 101 analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception:
Regarding limitation(s) (IV and V), under the broadest reasonable interpretation, recite steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions. Specifically, the courts have recognized computer functions directed to mere data gathering as well-understood, routine, and conventional functions when they are claimed in a merely generic manner or as insignificant extra-solution activity when considering evidence in view of Berkheimer v. HP, Inc., 881 F.3d 1360, 1368, 125 USPQ2d 1649, 1654 (Fed. Cir. 2018), see USPTO Berkheimer Memorandum (April 2018)).
Examiner uses Berkheimer: Option 2, a citation to one or more of the court decisions discussed in MPEP 2106.05(d)(II) as noting well-understood, routine, and conventional nature of the additional elements:
Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) (using a telephone for image transmission); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network). See MPEP 2106.05(d)(II).
Further, limitations (VI and VII), under the broadest reasonable interpretation, merely recite steps that apply generic computer components as a tool to perform judicial exceptions, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Regarding limitation (VIII), under the broadest reasonable interpretation, recite mere instructions to apply an exception, because they recite no more than an idea of a solution or outcome which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Considering additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 2, it is dependent upon claim 1 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 2 recites determining,…and based on the training dataset, a discretized data element set, wherein the discretized data element set comprises a plurality of discretized data elements;. Under the broadest reasonable interpretation, the limitations recite formatting data which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. Claim 2 also recites determining,…and based on the discretized data element set, a baseline;. Under the broadest reasonable interpretation, the limitations recite determining normal non-biased behavior which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. Claim 2 also recites and generating…a discretized path set, wherein the discretized path set comprises one or more paths from the baseline to the plurality of discretized data elements. Under the broadest reasonable interpretation, the limitations recite determining multiple potential biased behavior which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. The steps of observation, evaluation, and judgement are mental processes. Claim 2 also recites by the bias identification engine. Under the broadest reasonable interpretation, the limitations merely recite steps that apply generic computer components to perform a judicial exception, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Therefore, claim 2 does not solve the deficiencies of claim 1.
Regarding claim 3, it is dependent upon claim 2 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 3 recites generating,…and using the UDIG technique, an attribution score for each of the plurality of discretized data elements;. Under the broadest reasonable interpretation, the limitations recite using an integrated gradient which is interpreted as using a mathematical calculation. A mathematical calculation is interpreted as a mathematical concept. Claim 3 also recites and determining,…an individual bias score for each data element included in the discretized data element set, wherein the individual bias score for each of the plurality of discretized data elements is based on its corresponding attribution score. Under the broadest reasonable interpretation, the limitations recite determining bias scores which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. The steps of observation, evaluation, and judgement are mental processes. Claim 3 also recites by the bias identification engine. Under the broadest reasonable interpretation, the limitations merely recite steps that apply generic computer components to perform a judicial exception, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Therefore, claim 3 does not solve the deficiencies of claim 2.
Regarding claim 4, it is dependent upon claim 3 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 4 recites further comprising: comparing…the individual bias score for each of the plurality of discretized data elements to a bias identification threshold;. Under the broadest reasonable interpretation, the limitations recite comparing a score to a threshold which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. Claim 4 also recites and determining,…and based on the comparing, whether a data element corresponds to a biased data element. Under the broadest reasonable interpretation, the limitations recite determining whether a score satisfies a threshold which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. The steps of observation, evaluation, and judgement are mental processes. Claim 4 also recites by the bias identification engine. Under the broadest reasonable interpretation, the limitations merely recite steps that apply generic computer components to perform a judicial exception, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Therefore, claim 4 does not solve the deficiencies of claim 3.
Regarding claim 5, it is dependent upon claim 3 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 5 recites wherein the UDIG technique is periodically applied to the discretized data element set during the model training session. Under the broadest reasonable interpretation, the limitations recite using an integrated gradient calculation which is interpreted as using a mathematical calculation. A mathematical calculation is interpreted as a mathematical concept. Therefore, claim 5 does not solve the deficiencies of claim 3.
Regarding claim 6, it is dependent upon claim 1 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 6 recites wherein causing performance of the bias mitigation action comprises: applying,…and based on the bias identification event type, a soft lock to the training dataset;. Under the broadest reasonable interpretation, the limitations recite locking a dataset which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. Claim 6 also recites determining,…and based on the bias identification event type, a bias removal technique;. Under the broadest reasonable interpretation, the limitations recite determining how to remove a bias data element which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. Claim 6 also recites applying…the bias removal technique to the training dataset;. Under the broadest reasonable interpretation, the limitations recite removing biased data elements which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. Claim 6 also recites determining…whether the training dataset comprises the biased data;. Under the broadest reasonable interpretation, the limitations recite identifying if there is biased data which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. The steps of observation, evaluation, and judgement are mental processes. Claim 6 also recites and in an instance in which the bias removal technique removed the biased data, storing, by the bias treatment circuitry, the training dataset in a storage device. Under the broadest reasonable interpretation, the limitations recite steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions. Specifically, the courts have recognized computer functions directed to mere data gathering as well-understood, routine, and conventional functions when they are claimed in a merely generic manner or as insignificant extra-solution activity (MPEP 2106.05(g)). Claim 6 also recites by the bias treatment circuitry. Under the broadest reasonable interpretation, the limitations merely recite steps that apply generic computer components to perform a judicial exception, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Therefore, claim 6 does not solve the deficiencies of claim 1.
Regarding claim 7, it is dependent upon claim 6 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 7 recites further comprising: in an instance in which the training dataset comprises the biased data, maintaining…the soft lock. Under the broadest reasonable interpretation, the limitations recite determining whether to use a lock based on a criteria which is a step of observation, evaluation, and judgement which can be performed mentally or with pen and paper. The steps of observation, evaluation, and judgement are mental processes. Claim 7 also recites by the bias treatment circuitry. Under the broadest reasonable interpretation, the limitations merely recite steps that apply generic computer components to perform a judicial exception, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Therefore, claim 7 does not solve the deficiencies of claim 6.
Regarding claim 8, it is dependent upon claim 1 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 8 recites wherein the machine learning model is a Large Language Model (LLM). Under the broadest reasonable interpretation, the limitations merely recite steps that apply a generic LLM, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Therefore, claim 8 does not solve the deficiencies of claim 1.
Regarding claim 9, in step 1 of the 101 analysis set forth in MPEP 2106, the claim recites An apparatus for dynamically identifying bias in a dataset, the apparatus comprising: communications hardware. The claim recites an apparatus which is interpreted as a machine. A machine is one of the four statutory categories of invention. For the Step 2A/2B analyses, since claim 9 is similar to claim 1 it is rejected under the same rationales as claim 1.
The additional limitation below fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception.
the apparatus comprising: communications hardware configured to… (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
Considering additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claims 10-16, the claims are similar to claims 2-8 and are rejected under the same rationales.
Regarding claim 17, in step 1 of the 101 analysis set forth in MPEP 2106, the claim recites A computer program product for dynamically identifying bias in a dataset, the computer program product comprising a non-transitory computer-readable storage medium. The claim recites a non-transitory computer readable medium which is interpreted as an article of manufacture. An article of manufacture is one of the four statutory categories of invention. For the Step 2A/2B analyses, since claim 17 is similar to claim 1 it is rejected under the same rationales as claim 1.
The additional limitation below fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception.
A computer program product for dynamically identifying bias in a dataset, the computer program product comprising a non-transitory computer-readable storage medium storing instructions that, when executed by an apparatus, cause the apparatus to:… (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
Considering additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claims 18-20, the claims are similar to claims 2-4 and are rejected under the same rationales.
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 1-2, 9-10, and 17-18 are rejected under 35 U.S.C. 103 as being unpatentable over Rawat, et al., US Pre-Grant Publication US20230289573A1 (“Rawat”) in view of Sanyal, et al., Non-Patent Literature “Discretized Integrated Gradients for Explaining Language Models” (“Sanyal”).
Regarding claim 1, Rawat discloses:
A method for dynamically identifying bias in a dataset, (Rawat, ⁋2, “However, deep generative models are often trained on limited data and require computational resource and expertise. On the other hand, deep generative models sourced from third parties may not produce data that obeys fairness criteria [A method for dynamically identifying bias in a dataset,].”).
the method comprising: receiving, by communications hardware, a fine-tuning request, wherein the fine-tuning request comprises a set of fine-tuning parameters; retrieving, by a bias identification engine and based on the set of fine-tuning parameters, a machine learning model and a training dataset, wherein the training dataset comprises a plurality of data elements; (Rawat, ⁋14, “System 100 [the method comprising: receiving, by communications hardware,] for assessing fairness of a deep generative model includes fairness assessor 110. From input 130, fairness assessor 110 [retrieving, by a bias identification engine] receives user defined fairness criterion 140 [a fine-tuning request, wherein the fine-tuning request comprises a set of fine-tuning parameters;] and pretrained deep generative model 150. Pretrained deep generative model 150 can generate synthetic samples [and based on the set of fine-tuning parameters, a machine learning model and a training dataset, wherein the training dataset comprises a plurality of data elements;].”).
during a model training session, determining, by the bias identification engine…, that a data element corresponds to biased data; (Rawat, ⁋19, “Fairness assessor 110 [during a model training session, determining, by the bias identification engine…,] assesses the fairness of pretrained deep generative model 150, as per user defined fairness criterion 140. For instance, fairness assessor 110 samples from pretrained deep generative model 150; fairness assessor 110 uses a classifier to evaluate the fairness or explore the latent space to evaluate fairness [that a data element corresponds to biased data;].”).
in response to determining that the data element corresponds to biased data, determining, by the bias identification engine, a bias identification event, wherein the bias identification event corresponds to a bias identification event type and the bias identification event type is based on the data element that corresponds to the biased data; (Rawat, ⁋21, “Often, pretrained deep generative model 150 incorporates a structured input where a part of the input is fed via a random seed while the other part is defined by attributes (e.g., “smile” or “glasses” in human portraits). In such cases, other techniques can be used for fairness assessment. For example, fairness assessor 110 [in response to determining that the data element corresponds to biased data, determining, by the bias identification engine,] can assess group fairness with respect to each of these attributes. Given structured inputs, fairness assessor 110 can also compare the output of pretrained deep generative model 150 for two different conditional inputs for a variety of random seeds; assessing fairness based on group attributes is interpreted as an event type (i.e. a bias identification event, wherein the bias identification event corresponds to a bias identification event type and the bias identification event type is based on the data element that corresponds to the biased data;).”).
determining, by bias treatment circuitry and based on the bias identification event type, a bias mitigation action; and causing, by the bias treatment circuitry, performance of the bias mitigation action. (Rawat, ⁋23, “System 100 [bias treatment circuitry] for assessing fairness of a deep generative model has output 160. Output 160 includes fairness report 170. Fairness report 170 is generated by fairness assessor 110. In fairness report 170, fairness assessor 110 summarizes the fairness assessment done by fairness assessor 110. Output 160 further includes recommendations 180 for modifying the deep generative model [determining, by bias treatment circuitry and based on the bias identification event type, a bias mitigation action; and causing, by the bias treatment circuitry, performance of the bias mitigation action.].”).
While Rawat teaches a bias mitigation strategy for generative models, Rawat does not explicitly teach:
…and using a Uniform Discretized Integrated Gradient (UDIG) technique…
Sanyal teaches …and using a Uniform Discretized Integrated Gradient (UDIG) technique… (Sanyal, abstract, “Here we propose Discretized Integrated Gradients (DIG), which allows effective attribution along non-linear interpolation paths. We develop two interpolation strategies for the discrete word embedding space that generates interpolation points that lie close to actual words in the embedding space, yielding more faithful gradient computation […and using a Uniform Discretized Integrated Gradient (UDIG) technique…].”).
Rawat and Sanyal are both in the same field of endeavor (i.e. machine learning). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Rawat and Sanyal to teach the above limitation(s). The motivation for doing so is that using discretized integrated gradients provides feature attribution for improved interpretability (cf. Sanyal, pg. 10295 col. 1, “Attribution-based explanations generate a scalar score for a given input feature that indicates the contribution (or importance) of that feature towards particular label (Ancona et al., 2018).”).
Regarding claim 2, Rawat in view of Sanyal teaches the method of claim 1. Sanyal further teaches:
further comprising: determining, by the bias identification engine and based on the training dataset, a discretized data element set, wherein the discretized data element set comprises a plurality of discretized data elements; (Sanyal, pg. 10288 col. 1, “Motivated by this, for a given input word embedding, we first search for an anchor word from the vocabulary that can be considered as the next interpolation point. Since the anchor point need not be monotonic w.r.t. the given input, we then optimally perturb the dimensions of the anchor word so that they satisfy the monotonicity constraints in Equation 2; perturbing the word embeddings to satisfy constraints is interpreted as discretizing or formatting the data (i.e. further comprising: determining, by the bias identification engine and based on the training dataset, a discretized data element set, wherein the discretized data element set comprises a plurality of discretized data elements;).”).
determining, by the bias identification engine and based on the discretized data element set, a baseline; (Sanyal, pg. 10288 col. 1, “Overall, given an initial input word embedding w and a baseline embedding w [determining, by the bias identification engine and based on the discretized data element set, a baseline;], our interpolation algorithm interpolates points from w to w (which is in decreasing order of k in Eq. 3).”).
and generating, by the bias identification engine, a discretized path set, wherein the discretized path set comprises one or more paths from the baseline to the plurality of discretized data elements. (Sanyal, pg. 10286 col. 2, “Below, we define our DIG formulation that allows interpolations along non-linear paths: xi DIG i(x) = xk i=xi ∂F xk ∂xi dxk i. (1) Here, xk i refers to the ith dimension of the kth interpolated point between input x and baseline x’ [and generating, by the bias identification engine, a discretized path set, wherein the discretized path set comprises one or more paths from the baseline to the plurality of discretized data elements.]”).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the present application to combine the teachings of Sanyal with the teachings of Rawat for the same reasons disclosed in claim 1.
Regarding claims 9-10, the claims are similar to claims 1-2 and are rejected under the same rationales.
Regarding claim 17, the claim is similar and is rejected under the same rationales as claim 1. Rawat further teaches the additional limitations A computer program product for dynamically identifying bias in a dataset, the computer program product comprising a non-transitory computer-readable storage medium storing instructions that, when executed by an apparatus, cause the apparatus to: (Rawat, claim 15, “A computer system for assessing fairness of a deep generative model, the computer system comprising one or more processors, one or more computer readable tangible storage devices, and program instructions stored on at least one of the one or more computer readable tangible storage devices for execution by at least one of the one or more processors, the program instructions executable to: [A computer program product for dynamically identifying bias in a dataset, the computer program product comprising a non-transitory computer-readable storage medium storing instructions that, when executed by an apparatus, cause the apparatus to:]”, and Rawat, ⁋41, “A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se [a non-transitory computer-readable storage medium]”).
Regarding claim 18, the claim is similar to claims 2 and is rejected under the same rationales.
Claims 3-5, 11-13, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Rawat, et al., US Pre-Grant Publication US20230289573A1 (“Rawat”) in view of Sanyal, et al., Non-Patent Literature “Discretized Integrated Gradients for Explaining Language Models” (“Sanyal”) and further in view of Dalli, et al., US Pre-Grant Publication US20220012591A1 (“Dalli”).
Regarding claim 3, Rawat in view of Sanyal teaches the method of claim 2. Sanyal further teaches further comprising: generating, by the bias identification engine and using the UDIG technique, an attribution score for each of the plurality of discretized data elements; (Sanyal, pg. 10287 col. 1, “Here, we describe our proposed interpolation algorithm that searches for intermediate interpolation points between the input word embedding and the baseline embedding. Once we have the desired interpolation points, we can use Equation 3 to compute the word attributions similar to the IG algorithm [further comprising: generating, by the bias identification engine and using the UDIG technique, an attribution score for each of the plurality of discretized data elements;].”).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the present application to combine the teachings of Sanyal with the teachings of Rawat for the same reasons disclosed in claim 1.
While the combination teaches a bias mitigation strategy using integrated gradients, the combination does not explicitly teach and determining, by the bias identification engine, an individual bias score for each data element included in the discretized data element set, wherein the individual bias score for each of the plurality of discretized data elements is based on its corresponding attribution score.
Dalli teaches and determining, by the bias identification engine, an individual bias score for each data element included in the discretized data element set, wherein the individual bias score for each of the plurality of discretized data elements is based on its corresponding attribution score. (Dalli, ⁋126, “Bias may be reported in the form of a feature importance graph or feature attribution graph highlighting the prominent features which are causing the unintentional bias [wherein the individual bias score for each of the plurality of discretized data elements is based on its corresponding attribution score.]. Error bars may also be used for demonstrating the impact on the model. This form of bias reporting may apply to all forms of datasets [and determining, by the bias identification engine, an individual bias score for each data element included in the discretized data element set,].”).
Rawat, in view of Sanyal, and Dalli are both in the same field of endeavor (i.e. bias mitigation). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Rawat, in view of Sanyal, and Dalli to teach the above limitation(s). The motivation for doing so is that feature attribution improves the detection of bias (cf. Dalli, ⁋47-49).
Regarding claim 4, Rawat in view of Sanyal and Dalli teaches the method of claim 3. Dalli further teaches further comprising: comparing, by the bias identification engine, the individual bias score for each of the plurality of discretized data elements to a bias identification threshold; and determining, by the bias identification engine and based on the comparing, whether a data element corresponds to a biased data element. (Dalli, ⁋48, “Bias detection may be detected by analysing the difference in feature attribution when applying a control swap [further comprising: comparing, by the bias identification engine, the individual bias score for each of the plurality of discretized data elements to a bias identification threshold; and determining, by the bias identification engine and based on the comparing, whether a data element corresponds to a biased data element.]. Control swap is a method whereby for a given transaction X, the potential data field which has a bias (for example gender), is modified to test any change in the outcome and explanation.”).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the present application to combine the teachings of Dalli with the teachings of Rawat and Sanyal for the same reasons disclosed in claim 3.
Regarding claim 5, Rawat in view of Sanyal and Dalli teaches the method of claim 3. Sanyal further teaches wherein the UDIG technique is periodically applied to the discretized data element set during the model training session. (Sanyal, abstract, “Here we propose Discretized Integrated Gradients (DIG), which allows effective attribution along non-linear interpolation paths; multiple paths are interpreted as performing the technique multiple times (i.e. wherein the UDIG technique is periodically applied to the discretized data element set during the model training session.).”).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the present application to combine the teachings of Sanyal with the teachings of Rawat and Dalli for the same reasons disclosed in claim 1.
Regarding claims 11-13, the claims are similar to claims 3-5 and are rejected under the same rationales.
Regarding claims 19-20, the claims are similar to claims 3-4 and are rejected under the same rationales.
Claims 6-7 and 14-15 are rejected under 35 U.S.C. 103 as being unpatentable over Rawat, et al., US Pre-Grant Publication US20230289573A1 (“Rawat”) in view of Sanyal, et al., Non-Patent Literature “Discretized Integrated Gradients for Explaining Language Models” (“Sanyal”) and further in view of Lee, et al., US Pre-Grant Publication US20210319333A1 (“Lee”).
Regarding claim 6, Rawat in view of Sanyal teaches the method of claim 1. Rawat further teaches:
determining, by the bias treatment circuitry and based on the bias identification event type, a bias removal technique; (Rawat, ⁋23, “System 100 for assessing fairness of a deep generative model has output 160. Output 160 includes fairness report 170. Fairness report 170 is generated by fairness assessor 110. In fairness report 170, fairness assessor 110 summarizes the fairness assessment done by fairness assessor 110. Output 160 further includes recommendations 180 for modifying the deep generative model [determining, by the bias treatment circuitry and based on the bias identification event type, a bias removal technique;].”).
applying, by the bias treatment circuitry, the bias removal technique to the training dataset; (Rawat, ⁋24, “In one embodiment, deep generative model modifier 120 recommends to reweight inputs of pretrained deep generative model 150. The inputs to pretrained deep generative model 150 provide one means to modify the output distributions. One simple strategy to ensure fairness is to reconfigure the input distribution as per user defined fairness criterion 140 [applying, by the bias treatment circuitry, the bias removal technique to the training dataset;].”).
determining, by the bias treatment circuitry, whether the training dataset comprises the biased data; (Rawat, ⁋21, “Often, pretrained deep generative model 150 incorporates a structured input where a part of the input is fed via a random seed while the other part is defined by attributes (e.g., “smile” or “glasses” in human portraits). In such cases, other techniques can be used for fairness assessment. For example, fairness assessor 110 can assess group fairness with respect to each of these attributes [determining, by the bias treatment circuitry, whether the training dataset comprises the biased data;].”).
While the combination teaches a bias mitigation strategy in training data, the combination does not explicitly teach:
wherein causing performance of the bias mitigation action comprises: applying, by the bias treatment circuitry and based on the bias identification event type, a soft lock to the training dataset;
and in an instance in which the bias removal technique removed the biased data, storing, by the bias treatment circuitry, the training dataset in a storage device.
Lee teaches:
wherein causing performance of the bias mitigation action comprises: applying, by the bias treatment circuitry and based on the bias identification event type, a soft lock to the training dataset; (Lee, ⁋120 and see Figure 9, “Once the predictive model is modified the process may terminate. Alternatively, the process may return to block 904 to determine if the modification reduces or eliminates the bias in the predictive model or to determine if the modification introduces a new bias; because the system checks if new bias was introduced after bias mitigation, this is interpreted as putting a soft lock on the dataset to process the dataset again (i.e. applying, by the bias treatment circuitry and based on the bias identification event type, a soft lock to the training dataset;).”).
and in an instance in which the bias removal technique removed the biased data, storing, by the bias treatment circuitry, the training dataset in a storage device. (Lee, ⁋118 and see Figure 9, “At block 932, the computing device generate modification to the training dataset to remove the bias [and in an instance in which the bias removal technique removed the biased data,].” and Lee, ⁋120, “Once the predictive model is modified the process may terminate; the process in Figure 9 terminating is interpreted as the training dataset having the bias removed and thus saved for later processing (i.e. storing, by the bias treatment circuitry, the training dataset in a storage device.). Alternatively, the process may return to block 904 to determine if the modification reduces or eliminates the bias in the predictive model or to determine if the modification introduces a new bias.”).
Rawat, in view of Sanyal, and Lee are both in the same field of endeavor (i.e. bias mitigation). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Rawat, in view of Sanyal, and Lee to teach the above limitation(s). The motivation for doing so is that bias mitigation might need to be repeated due to introducing new bias during mitigation (cf. Lee, ⁋120, “Once the predictive model is modified the process may terminate. Alternatively, the process may return to block 904 to determine if the modification reduces or eliminates the bias in the predictive model or to determine if the modification introduces a new bias.”).
Regarding claim 7, Rawat in view of Sanyal and Lee teaches the method of claim 6. Lee further teaches further comprising: in an instance in which the training dataset comprises the biased data, maintaining, by the bias treatment circuitry, the soft lock. (Lee, ⁋120 and see Figure 9, “Once the predictive model is modified the process may terminate. Alternatively, the process may return to block 904 to determine if the modification reduces or eliminates the bias in the predictive model or to determine if the modification introduces a new bias; because the system checks if new bias was introduced after bias mitigation, this is interpreted as putting a soft lock on the dataset to process the dataset again (i.e. further comprising: in an instance in which the training dataset comprises the biased data, maintaining, by the bias treatment circuitry, the soft lock.).”).
It would have been obvious to one of ordinary skill in the art before the effective filling date of the present application to combine the teachings of Lee with the teachings of Rawat and Sanyal for the same reasons disclosed in claim 6.
Regarding claims 14-15, the claims are similar to claims 6-7 and are rejected under the same rationales.
Claims 8 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Rawat, et al., US Pre-Grant Publication US20230289573A1 (“Rawat”) in view of Sanyal, et al., Non-Patent Literature “Discretized Integrated Gradients for Explaining Language Models” (“Sanyal”) and further in view of Lee, et al., Non-Patent Literature “What Are Large Language Models Used For?” (“Lee2”).
Regarding claim 8, Rawat in view of Sanyal teaches the method of claim 1. While the combination teaches a bias mitigation strategy for generative models, the combination does not explicitly teach wherein the machine learning model is a Large Language Model (LLM).
Lee2 teaches wherein the machine learning model is a Large Language Model (LLM). (Lee2, pg. 1, “A large language model, or LLM, is a deep learning algorithm that can recognize, summarize, translate, predict and generate text and other forms of content based on knowledge gained from massive datasets [wherein the machine learning model is a Large Language Model (LLM).].”).
Rawat, in view of Sanyal, and Lee2 are both in the same field of endeavor (i.e. generative models). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Rawat, in view of Sanyal, and Lee2 to teach the above limitation(s). The motivation for doing so is that using LLMs improves the complexity and capabilities of artificial intelligence applications (cf. Lee2, pg. 2, “These models broaden AI’s reach across industries and enterprises, and are expected to enable a new wave of research, creativity and productivity, as they can help to generate complex solutions for the world’s toughest problems.”).
Regarding claim 16, the claim is similar to claim 8 and is rejected under the same rationales.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Tran, et al., US20250322293A1 discloses a system that identifies and mitigates bias at three different stages of a machine learning pipeline.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NICHOLAS S WU whose telephone number is (571)270-0939. The examiner can normally be reached Monday - Friday 8:00 am - 4:00 pm 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, Michelle Bechtold can be reached at 571-431-0762. 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.
/N.S.W./Examiner, Art Unit 2148 /MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
Prosecution Insights