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 .
DETAILED ACTION
Claims 1-20 are pending. Claims 1, 8, and 15 are independent.
This Application was published as US 20250307532.
A parent priority date is 19 May 2022.
The instant Application is directed to a method of generating a final model by ensembling trained models.
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 a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more. The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Step 1: The independent Claims are directed to statutory categories:
Claim 1 is a method claim and directed to the process category of patentable subject matter.
Claim 8 is a device claim and directed to the machine or manufacture category of patentable subject matter.
Claim 15 is a non-transitory machine-readable medium claim and is directed to the machine or manufacture category of patentable subject matter.
Step 2A, Prong One: Does the Claim recite a Judicially Recognized Exception? Abstract Idea? Are these Claims nevertheless considered Abstract as a Mathematical Concept (mathematical relationships, mathematical formulas or equations, mathematical calculations), Mental Process (concepts performed in the human mind (including an observation, evaluation, judgment, opinion), or Certain Methods of Organizing Human Activity (1-fundamental economic principles or practices (including hedging, insurance, mitigating risk), 2-commercial or legal interactions (including agreements in the form of contracts; legal obligations; advertising, marketing or sales activities or behaviors; business relations), 3- managing personal behavior or relationships or interactions between people (including social activities, teaching, and following rules or instructions) and fall under the judicial exception to patentable subject matter?)
The rejected Claims recite Mathematical Concepts or Mental Processes.
Step 2A, Prong Two: Additional Elements that Integrate the Judicial Exception into a Practical Application? Identifying whether there are any additional elements recited in the claim beyond the judicial exception(s), and evaluating those additional elements to determine whether they integrate the exception into a practical application of the exception. “Integration into a practical application” requires an additional element(s) or a combination of additional elements in the claim to apply, rely on, or use the judicial exception in a manner that imposes a meaningful limit on the judicial exception, such that the claim is more than a drafting effort designed to monopolize the exception. Uses the considerations laid out by the Supreme Court and the Federal Circuit to evaluate whether the judicial exception is integrated into a practical application.
The rejected Claims do not include additional limitations that point to integration of the abstract idea into a practical application and are therefore directed to a Mental Process or a Mathematical Concept.
Claim 1 is a generic automation of a mental process because a human agent can receive a training dataset and update a model via a user interface, and then use the model to process a document. Prong Two of step 2A in the 101 analysis asks whether the abstract idea is integrated with a practical application. The answer is no in this instance because there is no technological solution in the Claim that “integrates” the abstract idea. The Claim only suggests that the abstract idea be applied. It does not describe an application.
1. A method for training a machine learning model, the method comprising:
receiving a training dataset; (a human can receive a dataset at a user interface)
generating an anti-expert model by updating weights of a base model based on a training objective utilizing the training dataset; (a human can perform steps to train a model at a user interface)
generating a final model based at least in part on computing weights of the final model by subtracting weights of the anti-expert model from weights of the base model; (a human can perform steps to subtract weights; subtraction of weights is also a mathematical concept)
receiving, via a user interface, a user input document; and automatically generating, by the generated final model, an output presented at the user interface in response to the user input document. (a user can perform steps to process a document with a model)
Step 2B: Search for Inventive Concept: Additional Element Do not amount to Significantly More: The limitations of "model" and “user interface” are well-understood, routine, and conventional machine components that are being used for their well-understood, routine, and conventional and rather generic functions. Additionally, these limitations are expressed parenthetically and lack nexus to the Claim language and as such are a separable and divisible mention to a machine. Accordingly, they are not sufficient to cause the Claim to amount to significantly more than the underlying abstract idea.
The Dependent Claims do not add limitations that could help the Claim as a whole to amount to significantly more than the Abstract idea identified for the Independent Claim:
2. The method of claim 1, further comprising: determining a first subset of data from a source dataset based on data in the first subset meeting a threshold value of a metric, wherein the training dataset includes the first subset of data. (a user can sort data manually by comparing it to a threshold)
3. The method of claim 1, further comprising: generating an expert model by updating weights of the base model based on a second training dataset, wherein generating the final model is further based on weights computed by adding weights of the expert model to the weights of the base model. (a user can initiate training a model on a specific dataset)
4. The method of claim 3, wherein the weights of the expert model, and the weights of the anti-expert model are scaled using respective mixing coefficients. (mixing coefficients are mathematical concepts)
5. The method of claim 3, further comprising: determining a first subset of data from a source dataset based on data in the first subset meeting a threshold value of a metric; and determining a second subset of data from the source dataset based on data in the second subset not meeting the threshold value of the metric, wherein the training dataset includes the first subset of data and the second training dataset includes the second subset of data. (a user can determine two subsets of data based on a threshold)
6. The method of claim 1, wherein the output presented at the user interface in response to the user input document includes a summary of the user input document. (a user can input a document to a summary model)
7. The method of claim 1, wherein: the final model is a neural-network based model, and weights of the final model include weights of a neural network. (neural networks are well known models and in the claim it is used for a generic purpose)
The additional limitations introduced by the Dependent Claims are not sufficient as additional elements that integrate the judicial exception into a practical application or as additional elements that cause the Claim as a whole to amount to substantially more than the underlying abstract idea.
With respect to Independent Claim 8 and independent Claim 15, which have limitations similar to the limitations of Claim 1, the limitations of “memory,” “processor,” “communication interface,” and “machine-readable medium” are expressed parenthetically and lack nexus to the Claim language and as such are a separable and divisible mention to a machine. Accordingly, they do not include additional limitations that cause the Claim as a whole to amount to more than the underlying abstract idea.
The Dependent Claims 9-13 and 16-20 are similar to claims 2-6, and claim 14 is similar to claim 7, and they do not add limitations that could integrate the judicial exception into a practical application or help the Claim as a whole to amount to significantly more than the Abstract idea identified for the Independent Claim.
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is appropriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP § 2146 et seq. for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The filing of a terminal disclaimer by itself is not a complete reply to a nonstatutory double patenting (NSDP) rejection. A complete reply requires that the terminal disclaimer be accompanied by a reply requesting reconsideration of the prior Office action. Even where the NSDP rejection is provisional the reply must be complete. See MPEP § 804, subsection I.B.1. For a reply to a non-final Office action, see 37 CFR 1.111(a). For a reply to final Office action, see 37 CFR 1.113(c). A request for reconsideration while not provided for in 37 CFR 1.113(c) may be filed after final for consideration. See MPEP §§ 706.07(e) and 714.13.
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The actual filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/apply/applying-online/eterminal-disclaimer.
Claims 1-4, 6, 8-11, 13, 15-18, and 20 rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1, 3, 5, 7, 8, 10, 12, 14, 15, 17, 19, and 20 of U.S. Patent No. 12361201. Although the claims at issue are not identical, they are not patentably distinct from each other as mapped below.
Clm.
Element
Ref
Clm.
Reference Element (US 12361201 B2)
1
1. A method for training a machine learning model, the method comprising:
1
1. A method for training a summarization model, the method comprising:
1
receiving a training dataset;
1
receiving a training dataset,
1
generating an anti-expert model by updating weights of a base model based on a training objective utilizing the training dataset;
1
generating an anti-expert summarization model by updating parameters of the base summarization model based on a second training objective comparing summaries generated by the base summarization model and corresponding summaries from the noisy training dataset;
1
generating a final model based at least in part on computing weights of the final model by subtracting weights of the anti-expert model from weights of the base model;
1
generating a final summarization model based at least in part on parameters computed by subtracting weights of the anti-expert summarization model from weights of the base summarization model;
1
receiving, via a user interface, a user input document; and automatically generating, by the generated final model, an output presented at the user interface in response to the user input document.
1
receiving, via a user interface, a user input document; andautomatically generating, by the generated final summarization model, an output summary presented at the user interface in response to the user input document.
2
2. The method of claim 1, further comprising: determining a first subset of data from a source dataset based on data in the first subset meeting a threshold value of a metric, wherein the training dataset includes the first subset of data.
3
determining a score for each document and corresponding summary based on the first factual metric; andusing the score to select those which meet a predetermined threshold.
3
3. The method of claim 1, further comprising: generating an expert model by updating weights of the base model based on a second training dataset, wherein generating the final model is further based on weights computed by adding weights of the expert model to the weights of the base model.
5
generating an expert summarization model by updating parameters of the base summarization model based on comparing summaries generated by the base summarization model and corresponding summaries from the clean training dataset;wherein generating the final summarization model is further based on parameters computed by adding parameters of the expert summarization model to the parameters of the base summarization model.
4
4. The method of claim 3, wherein the weights of the expert model, and the weights of the anti-expert model are scaled using respective mixing coefficients.
7
7. The method of claim 5, wherein the parameters of the base summarization model, the parameters of the expert summarization model, and the parameters of the anti-expert summarization model are scaled using respective mixing coefficients.
6
6. The method of claim 1, wherein the output presented at the user interface in response to the user input document includes a summary of the user input document.
1
automatically generating, by the generated final summarization model, an output summary presented at the user interface in response to the user input document.
8
8. A system for training a machine learning model, the system comprising:
8
8. A system for training a summarization model, the system comprising:
8
a memory that stores a base model and a plurality of processor executable instructions;
8
a memory that stores a summarization model and a plurality of processor executable instructions;
8
a communication interface that receives a training dataset; and
8
a communication interface that receives a training dataset, the training dataset comprising a plurality of documents and a plurality of summaries corresponding to the plurality of documents; and
8
one or more hardware processors that read and execute the plurality of processor- executable instructions from the memory to perform operations comprising:
8
one or more hardware processors that read and execute the plurality of processor-executable instructions from the memory to perform operations comprising:
8
generating an anti-expert model by updating weights of the base model based on a training objective utilizing the training dataset;
8
generating an anti-expert summarization model by updating parameters of the base summarization model based on a second training objective comparing summaries generated by the base summarization model and corresponding summaries from the noisy training dataset;
8
generating a final model based at least in part on computing weights of the final model by subtracting weights of the anti-expert model from weights of the base model;
8
generating the summarization model based at least in part on parameters computed by subtracting weights of the anti-expert summarization model from weights of the base summarization model;
8
receiving, via a user interface, a user input document; and automatically generating, by the generated final model, an output presented at the user interface in response to the user input document.
8
receiving, via a user interface, a user input document; andautomatically generating, by the summarization model, an output summary presented at the user interface in response to the user input document.
9
9. The system of claim 8, wherein the one or more hardware processors perform operations further comprising: determining a first subset of data from a source dataset based on data in the first subset meeting a threshold value of a metric, wherein the training dataset includes the first subset of data.
10
determining a score for each document and corresponding summary based on the first factual metric; andusing the score to select those which meet a predetermined threshold.
10
10. The system of claim 8, wherein the one or more hardware processors perform operations further comprising: generating an expert model by updating weights of the base model based on a second training dataset, wherein generating the final model is further based on weights computed by adding weights of the expert model to the weights of the base model.
12
generating an expert summarization model by updating parameters of the base summarization model based on comparing summaries generated by the base summarization model and corresponding summaries from the clean training dataset; andgenerating the summarization model further based on parameters computed by adding parameters of the expert summarization model to the parameters of the base summarization model.
11
11. The system of claim 10, wherein the weights of the expert model, and the weights of the anti-expert model are scaled using respective mixing coefficients.
14
14. The system of claim 12, wherein the parameters of the base summarization model, the parameters of the expert summarization model, and the parameters of the anti-expert summarization model are scaled using respective mixing coefficients.
13
13. The system of claim 8, wherein the output presented at the user interface in response to the user input document includes a summary of the user input document.
8
automatically generating, by the summarization model, an output summary presented at the user interface in response to the user input document.
15
15. A non-transitory machine-readable medium comprising a plurality of machine-executable instructions which, when executed by one or more processors, are adapted to cause the one or more processors to perform operations comprising:
15
15. A non-transitory machine-readable medium comprising a plurality of machine-executable instructions which, when executed by one or more processors, are adapted to cause the one or more processors to perform operations comprising:
15
receiving a training dataset;
15
receiving a training dataset,
15
generating an anti-expert model by updating weights of a base model based on a training objective utilizing the training dataset;
15
generating an anti-expert summarization model by updating parameters of the base summarization model based on a second training objective comparing summaries generated by the base summarization model and corresponding summaries from the noisy training dataset;
15
generating a final model based at least in part on computing weights of the final model by subtracting weights of the anti-expert model from weights of the base model;
15
generating a final summarization model based at least in part on parameters computed by subtracting weights of the anti-expert summarization model from weights of the base summarization model;
15
receiving, via a user interface, a user input document; and automatically generating, by the generated final model, an output presented at the user interface in response to the user input document.
15
receiving, via a user interface, a user input document; andautomatically generating, by the generated final summarization model, an output summary presented at the user interface in response to the user input document.
16
16. The non-transitory machine-readable medium of claim 15, the plurality of machine-executable instructions further comprising: determining a first subset of data from a source dataset based on data in the first subset meeting a threshold value of a metric, wherein the training dataset includes the first subset of data.
17
determining a score for each document and corresponding summary based on the first factual metric; andusing the score to select those which meet a predetermined threshold.
17
17. The non-transitory machine-readable medium of claim 15, the plurality of machine-executable instructions further comprising: generating an expert model by updating weights of the base model based on a second training dataset, wherein generating the final model is further based on weights computed by adding weights of the expert model to the weights of the base model.
19
generating an expert summarization model by updating parameters of the base summarization model based on comparing summaries generated by the base summarization model and corresponding summaries from the clean training dataset;wherein generating the final summarization model is further based at least in part on parameters computed by adding parameters of the expert summarization model to the parameters of the base summarization model.
18
18. The non-transitory machine-readable medium of claim 17, wherein the weights of the expert model, and the weights of the anti-expert model are scaled using respective mixing coefficients.
20
20. The non-transitory machine-readable medium of claim 19, wherein the parameters of the base summarization model, the parameters of the expert summarization model, and the parameters of the anti-expert summarization model are scaled using respective mixing coefficients.
20
20. The non-transitory machine-readable medium of claim 15, wherein the output presented at the user interface in response to the user input document includes a summary of the user input document.
15
automatically generating, by the generated final summarization model, an output summary presented at the user interface in response to the user input document.
Claims 5, 7, 12, 14, and 19 rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1, 8, and 15 of U.S. Patent No. 12361201 in view of Werner et al. (US 20210406479 A1).
Regarding claim 5, claim 5 of US12361201B2 recites the limitations of claim 3 as mapped above, but not the additional limitations of claim 5.
Werner discloses: 5. The method of claim 3, further comprising: determining a first subset of data from a source dataset based on data in the first subset meeting a threshold value of a metric; ("[0020] The subsets of the training data 106 that include greater than (or equal to) a threshold number of OOV words or phrases (sometimes called OOV concepts) are considered high OOV training data 108…”)
and determining a second subset of data from the source dataset based on data in the second subset not meeting the threshold value of the metric, ("[0020] … The remaining subsets of the training data 106 can be considered low OOV training data 110. A threshold, this threshold or another threshold herein, can be determined based on a series of experiments that try to optimize system performance on a validation set. For example, an experiment can be performed to determine model accuracy based on using 2 or more OOV words or phrases as the threshold number. Another experiment can be performed to determine model accuracy based on using 3 or more OOV words or phrases as the threshold number and so on. Then the threshold that corresponds to the highest accuracy can be selected as the threshold. Similar experiments and selections can be made for other thresholds herein.")
wherein the training dataset includes the first subset of data and the second training dataset includes the second subset of data. (“[0022] The model trainer 104 can receive data representing a general purpose vocabulary model 112 and the high OOV training data 108 or the low OOV training data 110. The model trainer 104 can use transfer learning to generate a high OOV model 114. The high OOV model 114 can include the general purpose vocabulary model further trained based on the high OOV training data 108 (and not the low OOV training data 110).
[0023] The model trainer 104 can user transfer learning to generate a low OOV model 116. The low OOV model 114 can include the general purpose vocabulary model further trained based on the low OOV training data 108 (and not the high OOV training data 108).”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified US12361201B2 with choosing High OOV and Low OOV models based on different subsets of training data as taught by Werner. Doing so would have been beneficial to lower cost and time compared to a general purpose model, (Werner [0002]) overcome problems with general purpose domain models (Werner [0014]) and to optimize system performance (Werner [0020]).
Regarding claim 7, claim 1 of US12361201B2 recites the limitations of claim 1 as mapped above, but not the additional limitations of claim 7.
Werner discloses: 7. The method of claim 1, wherein: the final model is a neural-network based model, and weights of the final model include weights of a neural network. (“[0057] Some embodiments can be implemented using, at one or more operations, aspects of artificial intelligence (AI), including or processing and inferences performed using machine learning (ML) or neural networks (NNs) (e.g., deep NNs). AI is a field of technology concerned with developing decision-making systems to perform cognitive tasks that have traditionally required a living actor, such as a person. Neural networks (NNs) are computational structures that are loosely modeled on biological neurons. Generally, NNs encode information (e.g., data or decision making) via weighted connections (e.g., synapses) between nodes (e.g., neurons). Modern NNs are foundational to many AI applications.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified US12361201B2 with a neural network as taught by Werner. Neural networks are well known in the art, and are foundational to many AI applications (Werner [0057]).
Regarding claim 12, claim 14 of US12361201B2 recites the limitations of claim 10 as mapped above, but not the additional limitations of claim 12. Werner discloses the limitations of claim 12 as mapped in claim 5 above.
Regarding claim 14, claim 8 of US12361201B2 recites the limitations of claim 8 as mapped above, but not the additional limitations of claim 14. Werner discloses the limitations of claim 14 as mapped in claim 7 above.
Regarding claim 19, claim 20 of US12361201B2 recites the limitations of claim 18 as mapped above, but not the additional limitations of claim 19. Werner discloses the limitations of claim 19 as mapped in claim 5 above.
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)(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.
Claim(s) 1, 3, 6-8, 10, 13-15, 17, and 20 is/are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Chen et al. (US 20220383126 A1).
Regarding claim 1, Chen discloses: 1. A method for training a machine learning model, the method comprising: ("[0005] FIG. 1 is a block diagram illustrating the training of a dense layer of a model neural network to adapt a general model to a specific task or domain according to an example embodiment." )
receiving a training dataset; generating an anti-expert model by updating weights of a base model based on a training objective utilizing the training dataset; ("[0031] The first domain language model is trained at operation 230 with first domain specific training data without modifying base model weight matrices. Training may include the use of a loss function using standard backpropagation, calculating a gradient for every parameter and updating weights by subtracting the gradients." – the first domain language model reads on an anti-expert model because it is trained on different domain data.)
generating a final model based at least in part on computing weights of the final model by subtracting weights of the anti-expert model from weights of the base model; ("[0035] FIG. 4 is a flowchart illustrating a computer implemented method 400 of switching between domain models that utilize low-rank factorization matrices that have been combined as opposed to being used in parallel. Method 400 begins by removing the first low-rank factorization matrices at operation 410 by subtracting them from the combined base model weight matrices and corresponding first low-rank factorization matrices. At operation 420, second low-rank factorization matrices are added to the base model weight matrices. The second low-rank factorization matrices are treated as trainable parameters that are trained with second domain specific training data without modifying base model weight matrices.")
receiving, via a user interface, a user input document; and automatically generating, by the generated final model, an output presented at the user interface in response to the user input document. ("[0036] Method 400 may include performing inferencing on second domain language input using the base model weight matrices and corresponding second low-rank factorization matrices." – inferencing would be understood to include generating a response to a user document. [0053] discloses a user interface.)
Regarding claim 3, Chen discloses: 3. The method of claim 1, further comprising: generating an expert model by updating weights of the base model based on a second training dataset, (“[0035]… The second low-rank factorization matrices are treated as trainable parameters that are trained with second domain specific training data without modifying base model weight matrices.” – the second domain matrices read on an expert model.)
wherein generating the final model is further based on weights computed by adding weights of the expert model to the weights of the base model. ("[0035] FIG. 4 is a flowchart illustrating a computer implemented method 400 of switching between domain models that utilize low-rank factorization matrices that have been combined as opposed to being used in parallel. Method 400 begins by removing the first low-rank factorization matrices at operation 410 by subtracting them from the combined base model weight matrices and corresponding first low-rank factorization matrices. At operation 420, second low-rank factorization matrices are added to the base model weight matrices. The second low-rank factorization matrices are treated as trainable parameters that are trained with second domain specific training data without modifying base model weight matrices." )
Regarding claim 6, Chen discloses: 6. The method of claim 1, wherein the output presented at the user interface in response to the user input document includes a summary of the user input document. ("[0022] The general problem of adapting general domain models is now described to help illustrate the technical problems solved by the use of adaptation matrices. Consider adapting a pre-trained large-scale language model to conditional text generation tasks, such as summarization, machine reading comprehension (MRC), and natural language to SQL (NL2SQL), etc., where the training instances are context and target pairs: {(xi,yi)} i=1, ..." )
Regarding claim 7, Chen discloses: 7. The method of claim 1, wherein: the final model is a neural-network based model, and weights of the final model include weights of a neural network. ("[0004] A computer implemented method obtains neural network-based model base model weight matrices for each of multiple neural network layers. First low-rank factorization matrices are added to corresponding base model weight matrices to form a first domain model. The low-rank factorization matrices are treated as trainable parameters. The first domain model is trained with first domain specific training data without modifying base model weight matrices." )
Claim 8 is a system claim with limitations corresponding to the limitations of Claim 1 and is rejected under similar rationale. Additionally, a memory that stores a base model and a plurality of processor executable instructions; a communication interface that receives a training dataset; and one or more hardware processors of the Claim are taught by Chen (See fig. 6: Memory 603; Communication Interface 616; Processing unit 602)
Claim 10 is a system claim with limitations corresponding to the limitations of Claim 3 and is rejected under similar rationale.
Claim 13 is a system claim with limitations corresponding to the limitations of Claim 6 and is rejected under similar rationale.
Claim 14 is a system claim with limitations corresponding to the limitations of Claim 7 and is rejected under similar rationale.
Claim 15 is a medium claim with limitations corresponding to the limitations of Claim 1 and is rejected under similar rationale. Additionally, a non-transitory machine-readable medium comprising a plurality of machine-executable instructions of the Claim are taught by Chen (See fig. 6: Memory 603)
Claim 17 is a medium claim with limitations corresponding to the limitations of Claim 3 and is rejected under similar rationale.
Claim 20 is a medium claim with limitations corresponding to the limitations of Claim 6 and is rejected under similar rationale.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 2, 5, 9, 12, 16, and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chen in view of Werner et al. (US 20210406479 A1).
Regarding claim 2, Chen does not disclose the additional limitations.
Werner discloses: 2. The method of claim 1, further comprising: determining a first subset of data from a source dataset based on data in the first subset meeting a threshold value of a metric, wherein the training dataset includes the first subset of data. ("[0020] The subsets of the training data 106 that include greater than (or equal to) a threshold number of OOV words or phrases (sometimes called OOV concepts) are considered high OOV training data 108. The remaining subsets of the training data 106 can be considered low OOV training data 110. A threshold, this threshold or another threshold herein, can be determined based on a series of experiments that try to optimize system performance on a validation set. For example, an experiment can be performed to determine model accuracy based on using 2 or more OOV words or phrases as the threshold number. Another experiment can be performed to determine model accuracy based on using 3 or more OOV words or phrases as the threshold number and so on. Then the threshold that corresponds to the highest accuracy can be selected as the threshold. Similar experiments and selections can be made for other thresholds herein." – see also Fig. 1.)
Chen and Werner are considered analogous art to the claimed invention because they disclose training domain specific models. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Chen with choosing High OOV and Low OOV models based on different subsets of training data as taught by Werner. Doing so would have been beneficial to lower cost and time compared to a general purpose model, (Werner [0002]) overcome problems with general purpose domain models (Werner [0014]) and to optimize system performance (Werner [0020]).
Regarding claim 5, Chen does not disclose the additional limitations.
Werner discloses: 5. The method of claim 3, further comprising: determining a first subset of data from a source dataset based on data in the first subset meeting a threshold value of a metric; ("[0020] The subsets of the training data 106 that include greater than (or equal to) a threshold number of OOV words or phrases (sometimes called OOV concepts) are considered high OOV training data 108…”)
and determining a second subset of data from the source dataset based on data in the second subset not meeting the threshold value of the metric, ("[0020] … The remaining subsets of the training data 106 can be considered low OOV training data 110. A threshold, this threshold or another threshold herein, can be determined based on a series of experiments that try to optimize system performance on a validation set. For example, an experiment can be performed to determine model accuracy based on using 2 or more OOV words or phrases as the threshold number. Another experiment can be performed to determine model accuracy based on using 3 or more OOV words or phrases as the threshold number and so on. Then the threshold that corresponds to the highest accuracy can be selected as the threshold. Similar experiments and selections can be made for other thresholds herein.")
wherein the training dataset includes the first subset of data and the second training dataset includes the second subset of data. (“[0022] The model trainer 104 can receive data representing a general purpose vocabulary model 112 and the high OOV training data 108 or the low OOV training data 110. The model trainer 104 can use transfer learning to generate a high OOV model 114. The high OOV model 114 can include the general purpose vocabulary model further trained based on the high OOV training data 108 (and not the low OOV training data 110).
[0023] The model trainer 104 can user transfer learning to generate a low OOV model 116. The low OOV model 114 can include the general purpose vocabulary model further trained based on the low OOV training data 108 (and not the high OOV training data 108).”)
Chen and Werner are considered analogous art to the claimed invention because they disclose training domain specific models. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Chen with choosing High OOV and Low OOV models based on different subsets of training data as taught by Werner. Doing so would have been beneficial to lower cost and time compared to a general purpose model, (Werner [0002]) overcome problems with general purpose domain models (Werner [0014]) and to optimize system performance (Werner [0020]).
Claim 9 is a system claim with limitations corresponding to the limitations of Claim 2 and is rejected under similar rationale.
Claim 12 is a system claim with limitations corresponding to the limitations of Claim 5 and is rejected under similar rationale.
Claim 16 is a medium claim with limitations corresponding to the limitations of Claim 2 and is rejected under similar rationale.
Claim 19 is a medium claim with limitations corresponding to the limitations of Claim 5 and is rejected under similar rationale.
Claim(s) 4, 11, and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wortsman et al. ("Robust fine-tuning of zero-shot models" (Version 2)).
Regarding claim 4, Chen does not disclose the additional limitations.
Wortsman discloses: 4. The method of claim 3, wherein the weights of the expert model, and the weights of the anti-expert model are scaled using respective mixing coefficients. ("Step 2: Weight-space ensembling. For a mixing coefficient α ∈ [0,1], we consider the weight-space ensemble between the zero-shot model with parameters θ0 and the model obtained via standard fine-tuning with parameters θ1. The predictions of the weight-space ensemble wse are given by wse(x,α) = f(x,(1 −α)·θ0 +α·θ1) , (1)…" pg. 6, para 2)
Chen and Wortsman are considered analogous art to the claimed invention because they disclose training domain specific models. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Chen to ensemble models with a mixing coefficient as disclosed by Wortsman. Doing so would have been beneficial to combine models and achieve the best of both worlds (Wortsman pg. 5, para 5), and to retain high accuracy (Wortsman pg. 6, para 4).
Claim 11 is a system claim with limitations corresponding to the limitations of Claim 4 and is rejected under similar rationale.
Claim 18 is a medium claim with limitations corresponding to the limitations of Claim 4 and is rejected under similar rationale.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Wang et al. (US 20190347570 A1). Wang discloses a method of training a neural network by training with a first dataset of noisy data pairs and a second dataset of higher quality data pairs. See [0025] and Fig. 1.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JON C MEIS whose telephone number is (703)756-1566. The examiner can normally be reached Monday - Thursday, 8:30 am - 5:30 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, Hai Phan can be reached at 571-272-6338. 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.
/JON CHRISTOPHER MEIS/ Examiner, Art Unit 2654
/HAI PHAN/ Supervisory Patent Examiner, Art Unit 2654
Prosecution Insights