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 . This action is in response to the application filled on 08/29/2023. Claims 1-20 are pending and have been examined.
Priority
Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
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:
Step 1: The claim recites a method which falls into the statutory statue of process.
Step 2A Prong 1: The claim recites multiple abstract ideas: creating, by the computer program, a similarity matrix for the industry tags using a minimum labeling strategy, wherein the similarity matrix comprises a plurality of similarity scores for pairs of industry tags, a human being can look at pairs of labels, determine how similar they are, and then create a matrix, sampling, by the computer program, the industry tags using a stratified sampling method, a human being can pick labels using a stratified sampling method, generating, by the computer program, a semantic textual similarity style dataset comprising triplets of the industry tag descriptions, the company business descriptions, and the similarity scores, a human being can group these three things into triplets, inferring, by the computer program, an industry tag for each company using the checkpoint model, wherein the checkpoint model generates a cosine similarity for pairs for industry tag descriptions, a human being can assign an industry tag to a company.
Step 2A Prong 2: The claim does not integrate the abstract idea into a practical application since the additional elements of:
receiving, by a computer program executed by an electronic device, input data comprising company business descriptions, industry tags, and industry tag descriptions, is insignificant extra-solution activity: data gathering.
The computer program executed by an electronic device, is mere instructions to apply the abstract idea by a generic computer.
fine-tuning, by the computer program, a baseline language model for a semantic similarity model, is mere instructions to apply the abstract idea by a generic computer (a fine-tuned baseline language model).
training, by the computer program, the semantic similarity model by subjecting embeddings generated for pairs of the company business description and industry tag descriptions to a cosine similarity function, is mere instructions to apply the abstract idea by a generic computer (a semantic similarity model).
creating, by the computer program, a checkpoint model for the semantic similarity model
using the checkpoint model, wherein the checkpoint model generates a cosine similarity for pairs for industry tag descriptions, is mere instructions to apply the abstract idea by a generic computer (checkpoint model).
Step 2B: Claim 1 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above, receiving data is considered insignificant extra-solution activity because it is well-understood, routine, conventional activity as evidenced by MPEP §2106.05(d)(II)(I). Furthermore, the remaining additional elements amount to mere instructions to apply the judicial exception either by using a generic electronic device with a computer program, fine-tuning a generic language model without reciting details on how, training a generic semantic similarity model while only reciting specifics of the data used and not details on the training process, or creating and using a generic checkpoint model without reciting details of how the model is created. As such, the claim does not have any additional elements that amount to an integration of the judicial exception into a practical application nor to significantly more. Claim 1 is not patent eligible.
Regarding claim 2, the rejection of claim 1 is incorporated, further the claim recites evaluating, by the computer program, the checkpoint model using an Exact Match Ratio; and updating, by the computer program, the similarity matrix using cosine similarity values for the pairs of industry tag descriptions. Evaluating, by the computer program, the checkpoint model using an Exact Match Ratio, amounts to a mathematical concept given the evaluation is done using a mathematical formula involving mathematical calculations. Updating, by the computer program, the similarity matrix using cosine similarity values for the pairs of industry tag descriptions amounts to a mental process given a human being can update the values of pairs in a matrix. Claim 2 is not patent eligible.
Regarding claim 3, the rejection of claim 2 is incorporated, further the claim recites wherein the computer program evaluates the checkpoint model by comparing the Exact Match Ratio for the checkpoint model to a prior Exact Match Ratio for a prior model to determine improvement in the checkpoint model. This limitation amounts to a mental process given a human being can reasonably evaluate a model by comparing values from a prior model to determine if an improvement has occurred. Claim 3 is not patent eligible.
Regarding claim 4, the rejection of claim 3 is incorporated, further the claim recites optimizing, by the computer program, hyperparameters for the checkpoint model in response to the checkpoint model not improving relative to the prior model. This limitation amounts to a mental process given a human being can reasonably change the parameters of model after determining if it has improved using the aid of a generic computer to alter the values. Claim 4 is not patent eligible.
Regarding claim 5, the rejection of claim 1 is incorporated, further the claim recites receiving, by the computer program, feedback for the inferred industry tags. This limitation amounts to insignificant extra-solution activity: data gathering because it is well-understood, routine, conventional activity as evidenced by MPEP §2106.05(d)(II)(I). Claim 5 is not patent eligible.
Regarding claim 6, the rejection of claim 1 is incorporated, further the claim recites wherein the similarity scores are measured on a scale of between 0 and 5. This limitation amount to more specifics of the judicial exception given it merely specifies the range of values of the similarity scores. Claim 6 is not patent eligible.
Regarding claim 7, the rejection of claim 1 is incorporated, further the claim recites wherein the minimum labeling strategy receives between 10 percent and 15 percent of the similarity scores from subject matter experts. This limitation amounts to more specifics of the judicial exception given it merely specifies how much of the matrix is labeled by subject matter experts. Claim 7 is not patent eligible.
Regarding claim 8, the rejection of claim 1 is incorporated, further the claim recites wherein the stratified sampling method populates samples per similarity score such that each industry tag has a sample. This limitation amounts to more specifics of the judicial exception given it merely specifies how sampling is done by reciting a sampling method which can still be done mentally. Claim 8 is not patent eligible.
Regarding claim 9, the rejection of claim 1 is incorporated, further the claim recites wherein the baseline language model comprises a Robustly Optimized BERT Pre-training Approach model. This limitation amounts mere instructions to apply the judicial exception given it merely specifies the type of machine being used at a high level of generality to apply the judicial exception. Claim 9 is not patent eligible.
Regarding claim 10, the rejection of claim 1 Is incorporated, further the claim recites wherein the baseline language model is fine-tuned with text data that reference companies, industries, and/or industry taxonomies. This limitation amounts to merely linking the judicial exception to a field of use given It merely specifies the type of data being used to train the model. Claim 10 is not patent eligible.
Regarding claims 11-20, given the claims are merely variations of claims 1-10 in which a non-transitory computer readable storage medium is claimed instead of a method which falls into the statutory category of manufacture. Furthermore, the computer readable storage medium and processors are merely instructions to apply the judicial exception.
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) 1, 5, 9, 10, 11, 15, 19, 20 are rejected under 35 U.S.C. 103 as being unpatentable over Felgueiras (Creating Classification Models from Textual Descriptions of Companies Using Crunchbase) in view of Wu (Class2Simi: A Noise Reduction Perspective on Learning with Noisy Labels), Sechidis (On the Stratification of Multi-Label Data), and Reimers (Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks).
Regarding claim 1, Reimers teaches generating, by the computer program, a semantic textual similarity style dataset comprising triplets of the industry tag descriptions, the company business descriptions, and the similarity scores (Figure 1: SBERT architecture with classification objective function, creates (u, v, |u-v|) where |u-v| is the similarity and u and v are two sentences); fine-tuning, by the computer program, a baseline language model for a semantic similarity model (Section 2, In this publication, we use the pre-trained BERT and RoBERTa network and only fine-tune it to yield useful sentence embeddings. This reduces significantly the needed training time); training, by the computer program, the semantic similarity model by subjecting embeddings generated for pairs of the company business description and industry tag descriptions to a cosine similarity function (Section 3, In order to fine-tune BERT / RoBERTa, we create siamese and triplet networks (Schroff et al., 2015) to update the weights such that the produced sentence embeddings are semantically meaningful and can be compared with cosine-similarity……. The cosine similarity between the two sentence embeddings u and v is computed (Figure 2). We use mean squared-error loss as the objective function); creating, by the computer program, a checkpoint model for the semantic similarity model (Section 3, We fine-tune SBERT with a 3-way softmax classifier objective function for one epoch. We used a batch-size of 16, Adam optimizer with learning rate 2e−5, and a linear learning rate warm-up over 10% of the training data. Our default pooling strategy is MEAN, the fine-tuned model is the checkpoint model), and inferring, by the computer program, an industry tag for each company using the checkpoint model, wherein the checkpoint model generates a cosine similarity for pairs for industry tag descriptions (Section 4, At prediction time, we compute the cosine-similarity between the sentence embeddings, this can be applied to pairs of industry tag descriptions and company descriptions). Reimers fails to teach the specific context of industry and company matching and using similarity scores for pairs of labels.
Felgueiras teaches, which Reimers is silent on, receiving, by a computer program executed by an electronic device, input data comprising company business descriptions, industry tags, and industry tag descriptions (Section 3 Corpus, The extracted JSON entries contain a lot of information, but only a small portion of that information is relevant for our task: URL of the company, company name, description, short description, categories, and fine categories, fine categories are more specific and provide information about the category making them analogous to category descriptions).
Reimers and Felgueiras are analogous to the claimed invention because they discuss classification tasks. Therefore, it would have been obvious to one of ordinary skill in the art to have used the model setup, including the base-model, fine-tuning, and training, in Reimers alongside the specific context of company and industry matching found in Felgueiras to solve the classification problem in Felgueiras and efficiently find the similarity between sentences/descriptions (Reimers Abs).
The combination of Reimers and Felgueiras fails to teach creating, by the computer program, a similarity matrix for the industry tags using a minimum labeling strategy, wherein the similarity matrix comprises a plurality of similarity scores for pairs of industry tags.
Wu teaches creating, by the computer program, a similarity matrix for the industry tags (Figure 1, matrices of data pairs of labels) using a minimum labeling strategy (Section 4, Specifically, for each instance with clean label i, we replace its label by j with a probability of Tc,ij, the minimum labeling strategy is relabeling using the probability which changes the similarity matrix), wherein the similarity matrix comprises a plurality of similarity scores for pairs of industry tags (Section 3, if the two instances have the same class label, we assign this pair a similarity label 1, otherwise 0).
Wu is analogous to the claimed invention because it is in the field of multi-label classification. Therefore, it would have been obvious to one of ordinary skill in the art to have used the pairwise solution in Wu (including using the similarity between pairs of labels in the triplet) alongside the combination of Reimers and Felgueiras to solve the multi-class classification problem in Felgueiras because Wu’s method reduces noise rate in noisy labels (Wu Section 1, We theoretically prove that through this transformation, the noise rate becomes lower (see Theorem 2) an issue that may arise when doing multi-label classification.
The combination of Reimers, Felgueiras, and Wu fails to teach sampling, by the computer program, the industry tags using a stratified sampling method.
Sechidis teaches sampling, by the computer program, the industry tags using a stratified sampling method (Alg 2, We here propose an algorithm for achieving the relaxed version of multi-label stratification that we discussed in Sect. 2, goes on to show a stratification method for multi-class labeling in algorithm 2).
Sechidis is analogous to the claimed invention because it Is in the field of multi-label classification. Therefore, it would have been obvious to one of ordinary skill in the art to have used the stratification sampling methods used in Sechidis alongside the combination of Reimers, Felgueiras, and Wu to account for groups within a population of labels and maintain the proportion of these groups (Sechidis Abs).
Regarding claim 5, the method of claim 1 is taught by the combination above, further Felgueiras teaches further comprising: receiving, by the computer program, feedback for the inferred industry tags (Table 3 Multi-class classification results, provides feedback for the labeling by evaluating accuracy).
Regarding claim 9, the method of claim 1 is taught by the combination above, further Reimers teaches wherein the baseline language model comprises a Robustly Optimized BERT Pre-training Approach model (Section 2, In this publication, we use the pre-trained BERT and RoBERTa network and only fine-tune it to yield useful sentence embeddings. This reduces significantly the needed training time). The rationale for combination is the same as claim 1.
Regarding claim 10, the method of claim 1 is taught by the combination above, further Felgueiras teaches wherein the baseline language model is fine-tuned with text data that reference companies, industries, and/or industry taxonomies (Section 3, The extracted JSON entries contain a lot of information, but only a small portion of that information is relevant for our task: URL of the company, company name, description, short description, categories, and fine categories…..The final database contains a total of 405602 records, that have been randomly shuffled and stored into two different tables: train, containing 380602 records, will be used from training our models).
Regarding claims 11, 15, 19, and 20, the inventive concept is essentially the same as claims 1, 5, 9, and 10 with the addition of a non-transitory computer readable storage medium, including instructions stored thereon, which when read and executed by one or more computer processors, cause the one or more computer processors to perform steps which is implied by Reimers (Section 7, Performances were measured on a server with Intel i7-5820K CPU @ 3.30GHz, Nvidia Tesla V100 GPU, CUDA 9.2 and cuDNN. The results are depicted in Table 7).
Claim(s) 6, 16 are rejected under 35 U.S.C. 103 as being unpatentable over Felgueiras (Creating Classification Models from Textual Descriptions of Companies Using Crunchbase) in view of Wu (Class2Simi: A Noise Reduction Perspective on Learning with Noisy Labels), Sechidis (On the Stratification of Multi-Label Data), and Reimers (Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks) as applied to claims 1 and 11 above, and further in view of Cer (SemEval-2017 Task 1: Semantic Textual Similarity Multilingual and Cross-lingual Focused Evaluation).
Regarding claim 6, The combination of Reimers, Felgueiras, Wu, and Sechidis teaches the method of claim 1 but fails to teach wherein the similarity scores are measured on a scale of between 0 and 5. Cer teaches wherein the similarity scores are measured on a scale of between 0 and 5 (Table 1: Similarity scores with explanations and English examples from, Section 2, The ordinal scale in Table 1 guides human annotation, ranging from 0 for no meaning overlap to 5 for meaning equivalence).
Cer is analogous to the claimed invention because it is in the field of classification. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have used the similarity score scale in Cer along with the combination of Reimers, Felgueiras, Wu, and Sechidis because “The annotation scale is designed to be accessible by reasonable human judges without any formal expertise in linguistics” (Cer Section 2).
Regarding claim 16, The combination of Reimers, Felgueiras, Wu, and Sechidis teaches the non-transitory computer readable storage medium of claim 11 but fails to teach wherein the similarity scores are measured on a scale of between 0 and 5. Cer teaches wherein the similarity scores are measured on a scale of between 0 and 5 (Table 1: Similarity scores with explanations and English examples from, Section 2, The ordinal scale in Table 1 guides human annotation, ranging from 0 for no meaning overlap to 5 for meaning equivalence).
Cer is analogous to the claimed invention because it is in the field of classification. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have used the similarity score scale in Cer along with the combination of Reimers, Felgueiras, Wu, and Sechidis because “The annotation scale is designed to be accessible by reasonable human judges without any formal expertise in linguistics” (Cer Section 2).
Claim(s) 7, 17 are rejected under 35 U.S.C. 103 as being unpatentable over Felgueiras (Creating Classification Models from Textual Descriptions of Companies Using Crunchbase) in view of Wu (Class2Simi: A Noise Reduction Perspective on Learning with Noisy Labels), Sechidis (On the Stratification of Multi-Label Data), and Reimers (Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks) as applied to claims 1 and 11 above, and further in view of Maroo (US20210182606A1)
.
Regarding claim 7, The combination of Reimers, Felgueiras, Wu, and Sechidis teaches the method of claim 1 but fails to teach wherein the minimum labeling strategy receives between 10 percent and 15 percent of the similarity scores from subject matter experts. Maroo teaches wherein the minimum labeling strategy receives between 10 percent and 15 percent of the similarity scores from subject matter experts (Par 0014, This small labeled dataset is, for example, labeled by one or a small group of human experts so as to be highly accurate and consistent. For ease of reference herein, this small labeled dataset will be referred to as the “golden dataset.” The golden dataset is then applied to label a much larger training dataset for training a learning algorithm. In certain examples, the size of the golden dataset may be at least 2% of the size of the training dataset to be labeled. In other embodiments, the golden dataset may be larger compared to the training dataset to be labeled, such as 10% of the total data elements in the dataset to be labeled).
Maroo is analogous to the claimed invention because it is centered around labeling data for tasks such as classification. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have used the labeling strategy in Maroo along with the combination of Reimers, Felgueiras, Wu, and Sechidis to produce “high-quality labeled trading data—data points with reliably accurate labels—with minimal human involvement” (Maroo Par 0014).
Regarding claim 17, The combination of Reimers, Felgueiras, Wu, and Sechidis teaches the non-transitory computer readable storage medium of claim 11 but fails to teach wherein the minimum labeling strategy receives between 10 percent and 15 percent of the similarity scores from subject matter experts. Maroo teaches wherein the minimum labeling strategy receives between 10 percent and 15 percent of the similarity scores from subject matter experts (Par 0014, This small labeled dataset is, for example, labeled by one or a small group of human experts so as to be highly accurate and consistent. For ease of reference herein, this small labeled dataset will be referred to as the “golden dataset.” The golden dataset is then applied to label a much larger training dataset for training a learning algorithm. In certain examples, the size of the golden dataset may be at least 2% of the size of the training dataset to be labeled. In other embodiments, the golden dataset may be larger compared to the training dataset to be labeled, such as 10% of the total data elements in the dataset to be labeled).
Maroo is analogous to the claimed invention because it is centered around labeling data for tasks such as classification. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have used the labeling strategy in Maroo along with the combination of Reimers, Felgueiras, Wu, and Sechidis to produce “high-quality labeled trading data—data points with reliably accurate labels—with minimal human involvement” (Maroo Par 0014).
Claim(s) 8, 18 are rejected under 35 U.S.C. 103 as being unpatentable over Felgueiras (Creating Classification Models from Textual Descriptions of Companies Using Crunchbase) in view of Wu (Class2Simi: A Noise Reduction Perspective on Learning with Noisy Labels), Sechidis (On the Stratification of Multi-Label Data), and Reimers (Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks) as applied to claims 1 and 11 above, and further in view of Farias (Similarity Based Stratified Splitting: an approach to train better classifiers).
Regarding claim 8, the combination of Reimers, Felgueiras, Wu, and Sechidis teaches the method of claim 1 but fails to teach wherein the stratified sampling method populates samples per similarity score such that each industry tag has a sample. Farias teaches wherein the stratified sampling method populates samples per similarity score such that each industry tag has a sample (Algorithm 1 Algorithm for Similarity Based Stratified K-Fold Splitting, populates using similarity score for each label (industry tag)).
Farias is analogous to the claimed invention because it is centered around training classifiers for tasks like classification. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have used the stratification method in Farias along with the combination above “for a better representation of the data in the training phase” (Farias Abs).
Regarding claim 18, the inventive concept is essentially the same as claim 8 with the addition of a non-transitory computer readable storage medium, including instructions stored thereon, which when read and executed by one or more computer processors, cause the one or more computer processors to perform steps which is implied by Reimers (Section 7, Performances were measured on a server with Intel i7-5820K CPU @ 3.30GHz, Nvidia Tesla V100 GPU, CUDA 9.2 and cuDNN. The results are depicted in Table 7).
Claim(s) 2, 12 are rejected under 35 U.S.C. 103 as being unpatentable over Felgueiras (Creating Classification Models from Textual Descriptions of Companies Using Crunchbase) in view of Wu (Class2Simi: A Noise Reduction Perspective on Learning with Noisy Labels), Sechidis (On the Stratification of Multi-Label Data), and Reimers (Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks) as applied to claims 1 and 11 above, and further in view of Nguyen (Text Classification of Technical Papers Based on Text Segmentation).
Regarding claim 2, The combination of Reimers, Felgueiras, Wu, and Sechidis teaches the method of claim 1 and teaches updating, by the computer program, the similarity matrix using cosine similarity values for the pairs of industry tag descriptions (Wu Figure 2, An overview of the proposed method. We add a pairwise enumeration layer and similarity transition matrix to calculate and correct the predicted similarity posterior. By minimizing the proposed loss Lc2s, a classifier f can be learned for assigning clean labels, Section 1, To handle the transformed data pairs with noisy similarity labels, the connection between noisy similarity posterior and clean class posterior should be established. Intuitively, noisy similarity posterior can be linked to clean similarity posterior…. For the first part, we can draw on the philosophy of dealing with noisy class labels, e.g., selecting reliable data pairs for training, and correcting the similarity loss to learn a robust similarity classifier, noisy similarity labels, like those used in the matrix, are corrected which updates the similarity matrix, further the cosine measure of similarity can be used, as used in Reimers and claim 1 above). The combination fails to teach evaluating, by the computer program, the checkpoint model using an Exact Match Ratio.
Nguyen teaches evaluating, by the computer program, the checkpoint model using an Exact Match Ratio (Section 5, We used exact match ratio (EMR), accuracy, precision and recall as the instance-based metrics). Nguyen is analogous to the claimed invention because it discusses classification. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have used exact match ratio as a loss metric to evaluate models on classification tasks as done in Nguyen.
Regarding claim 12, the inventive concept is essentially the same as claim 2 with the addition of a non-transitory computer readable storage medium, including instructions stored thereon, which when read and executed by one or more computer processors, cause the one or more computer processors to perform steps which is implied by Reimers (Section 7, Performances were measured on a server with Intel i7-5820K CPU @ 3.30GHz, Nvidia Tesla V100 GPU, CUDA 9.2 and cuDNN. The results are depicted in Table 7).
Claim(s) 3, 4, 13, 14 are rejected under 35 U.S.C. 103 as being unpatentable over Felgueiras (Creating Classification Models from Textual Descriptions of Companies Using Crunchbase) in view of Wu (Class2Simi: A Noise Reduction Perspective on Learning with Noisy Labels), Sechidis (On the Stratification of Multi-Label Data), Reimers (Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks), and Nguyen (Text Classification of Technical Papers Based on Text Segmentation) as applied to claims 2 and 12 above, and further in view of Murphy (US20190325259A1).
Regarding claim 3, The combination of Reimers, Felgueiras, Wu, Nguyen, and Sechidis teaches the method of claim 2, but fails to teach wherein the computer program evaluates the checkpoint model by comparing the Exact Match Ratio for the checkpoint model to a prior Exact Match Ratio for a prior model to determine improvement in the checkpoint model. Murphy teaches wherein the computer program evaluates the checkpoint model by comparing the Exact Match Ratio for the checkpoint model to a prior Exact Match Ratio for a prior model to determine improvement in the checkpoint model (Par 0018, Comparing can include iteratively adjusting model parameters according to the algorithm to minimize loss, adjusting model parameters creates a new model which is compared with previous iterations model to minimize the loss, which can be the exact match ratio used in Nguyen to evaluate the model).
Murphy is analogous to the claimed invention because it is in the field of multi-label classification. Therefore, it would have been obvious to have used the iterative method in Murphy alongside the combination above to build a model that “minimizes loss” (Murphy Par 0018).
Regarding claim 4, The combination of Reimers, Felgueiras, Wu, Nguyen, Murphy, and Sechidis teaches the method of claim 3, further Murphy teaches optimizing, by the computer program, hyperparameters for the checkpoint model in response to the checkpoint model not improving relative to the prior model (Par 0057, Additionally, the system 100 can also adjust hyperparameters in the multi-label classification processor 112, until acceptable accuracy is realized.)
Regarding claims 13 and 14, the inventive concept is essentially the same as claims 3 and 4 with the addition of a non-transitory computer readable storage medium, including instructions stored thereon, which when read and executed by one or more computer processors, cause the one or more computer processors to perform steps which is implied by Reimers (Section 7, Performances were measured on a server with Intel i7-5820K CPU @ 3.30GHz, Nvidia Tesla V100 GPU, CUDA 9.2 and cuDNN. The results are depicted in Table 7).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NATNAEL A ASEGDEW whose telephone number is (571)270-0407. The examiner can normally be reached 7:30-5.
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, Kakali Chaki can be reached at (571) 272-3719. 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.
/NATNAEL A ASEGDEW/Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights