DETAILED ACTION
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 responsive to pending claims 1-20 filed 1/29/2024.
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.
Claim(s) 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The 35 U.S.C. 101 subject matter eligibility analysis first asks whether the claim is directed to one of the four statutory categories (Step 1). It next asks whether the claim is directed to an abstract idea (Step 2A), via Prong 1, whether an abstract idea (e.g., mathematical concept, mental process, certain methods of organizing human activity) is recited, and Prong 2, whether it is integrated into a practical application. It finally asks whether the claim as a whole includes additional elements that amount to significantly more than the judicial exception (Step 2B). See MPEP 2106.
STEP 1: The claims falls within one of the four statutory categories:
All claims are directed to methods and hardware processing systems and hence fall within one of the four statutory categories.
STEP 2A PRONG 1: The claims recite a judicial exception:
Claim 1 is directed to evaluating transaction data by consulting a token lookup, a mental process akin to what human examiner of transaction records may do. In particular (additional elements are underlined and subsequently analyzed):
For claim 1: a method performed by a processor, said method comprising:
tokenizing received transaction data to generate a plurality of tokens (tokenizing is just splitting up text and so can be compared mentally);
for the plurality of tokens, retrieving corresponding dictionary entries of predictability scores (retrieving scores for each token, such as via a lookup or via consideration, is a metal process) generated by a trained neural network, the neural network being trained on a labeled dataset with training tokens matched to corresponding vendors, the dictionary entries being generated based on prediction probabilities of the training tokens generated by the trained neural network (generating dictionary lookup entries of highly predictive tokens may be performed mentally);
calculating a maximum predictability score of the retrieved predictability scores (Judging a maximal value is a mental process);
providing the transaction data for downstream processing in response to determining that the maximum predictability score is above a threshold (making decisions based on a maximal calculation is a mental process); and
filtering out the transaction data in response to determining that the maximum predictability score is below the threshold (likewise, filtering or not considering data is a mental process).
For claim 2: The method of claim 1, the tokenizing the received transaction data further comprising:
receiving the transaction data comprising a transaction description; and
tokenizing the transaction description to generate the plurality of tokens (receiving a record and splitting it into tokens may be performed mentally).
For claim 3: The method of claim 1, the tokenizing the received transaction data further comprising:
receiving the transaction data comprising a transaction description; and
tokenizing the transaction description based on spaces within the transaction description to generate the plurality of tokens (receiving a record and splitting it into tokens may be performed mentally).
For claim 4: The method of claim 1, the providing the transaction data for downstream processing comprises:
providing the transaction data to an artificial intelligence pipeline (Provision may be performed mentally).
For claim 5: The method of claim 1, the retrieving the corresponding dictionary entries comprising:
retrieving the corresponding dictionary entries of the predictability scores (retrieving entries may be performed mentally) generated by a trained convolutional neural network.
For claim 6: The method of claim 1, further comprising:
assigning a predictability score of zero to a token that does not have a corresponding dictionary entry (Assigning of a zero value during consideration may be performed mentally).
For claim 7: The method of claim 1, the training of the neural network comprising:
generating a token-vendor matrix from the labeled dataset, the token-vendor matrix indicating counts of matches between tokens and corresponding vendors (generating a matrix or table may be performed mentally);
generating an updated token-vendor matrix by removing tokens with counts lower than a predetermined count threshold (removing table entries may be performed mentally); and
selecting the training tokens from the updated token-vendor matrix (performing selection may be performed mentally).
For claim 8: The method of claim 7, the selecting the training tokens comprising:
normalizing the counts of matches in the updated token-vendor matrix (performing normalization during consideration of records may be performed mentally); and
selecting the training tokens and corresponding normalized counts of matches in the updated token-vendor matrix (selection of tokens may be performed mentally).
For claim 9: The method of claim 7, the selecting the training tokens comprising:
normalizing the counts of matches in the updated token-vendor matrix (normalization during consideration of records may be performed mentally);
sorting the normalized counts of matches (sorting may be performed mentally);
removing the normalized counts of matches below a predetermined normalized counts of matches threshold (modifying tables, applying threshold criteria may be performed mentally); and
selecting the training tokens and corresponding normalized counts above the predetermined normalized counts of matches threshold (making selections based on criteria may be performed mentally).
For claim 10. The method of claim 1, further comprising:
retraining the neural network with a new labeled dataset (Performing retraining or updating a model may be performed mentally).
Claim 11 recites a system analogous to the above but additionally ecites the additional elements of: a non-transitory storage medium storing computer program instructions; and
a processor configured to execute the computer program instructions to cause operations.
STEP 2A PRONG 2: The claims do not integrate the exception into a practical application:
For claim 1, the additional elements recite generating by a trained neural network based on a labeled dataset comprising training tokens and matched vendors. However, the use of a neural network here is mere instructions to implement the abstract idea, the estimation of predictive value, on a neural networking device and hence does not constitute an integration into a practical application.
For claim 4, the additional elements recite provision to an artificial intelligence pipeline. However, the use of AI here is mere instructions to implement the abstract idea on an AI device and hence does not constitute an integration into a practical application.
For claim 5, 10, the additional elements recite generation or training by a CNN or a neural network. However, the use of CNN or neural networks here is mere instructions to implement the abstract idea on an AI device and hence does not constitute an integration into a practical application.
For claim 11, the additional elements recite non-transitory storage media and processing elements for executing the method. However, the use of a computer here is mere instructions to implement the mental process on a computing device and hence does not constitute an integration into a practical application.
STEP 2B: The claim as a whole do not include additional elements that amount to significantly more than the abstract idea:
For claim 1, the additional elements recite generating by a trained neural network based on a labeled dataset comprising training tokens and matched vendors. However, the use of a neural network for generating inferences of training tokens is well understood, routine, and conventional in the field of data analysis and hence does not constitute significantly more.
For claim 4, the additional elements recite provision to an artificial intelligence pipeline. However, the use of an AI pipeline for generating inferences of training tokens is well understood, routine, and conventional in the field of data analysis and hence does not constitute significantly more.
For claim 5, the additional elements recite generation or training by a CNN or neural network. However, the use of a CNN or neural network for generating inferences of training tokens is well understood, routine, and conventional in the field of data analysis and hence does not constitute significantly more.
For claim 11, the additional elements recite non-transitory storage media and processing elements for executing the method. However, the use of a computer for generating inferences of training tokens is well understood, routine, and conventional in the field of data analysis and hence does not constitute significantly more.
The remaining claims recite analogous computing systems and hence are rejected for the same reasons.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-4, 6-9, 11-14, 16-19 are rejected under 35 U.S.C. 103 as being unpatentable over Range (US 20200065710 A1) in view of Yeh ("Merchant category identification using credit card transactions", published 2020).
For claim 1, Range discloses: a method performed by a processor (fig.10:9010: processor, with figs.2-3 giving overview of the method), said method comprising:
tokenizing received transaction data to generate a plurality of tokens (fig.4:401, 0036);
for the plurality of tokens, retrieving corresponding dictionary entries of predictability scores (fig.2:220, 0029; fig.4:407-419, 0037-39: token correlation metric constitutes a predictability score, the scores being stored in some storage for use in later steps, hence, dictionary entries) generated by a trained machine learning model (0038 contemplates various machine learning techniques for determining predictive metric), the dictionary entries being generated based on prediction probabilities of the training tokens generated by the trained machine learning model (ibid: correlation metric is based prediction probability, with higher correlation corresponding to higher importance for some target value);
calculating a maximum predictability score of the retrieved predictability scores (0039 contemplates various thresholding for these predictive scores, hence, calculating maximum scores for determining thresholds);
providing the transaction data for downstream processing in response to determining that the maximum predictability score is above a threshold (ibid: higher than threshold scores are kept); and
filtering out the transaction data in response to determining that the maximum predictability score is below the threshold (ibid: transaction data represented by tokens is filtered from storage in response to not meeting threshold).
Range does not disclose: wherein the machine learning model is a neural network, the neural network being trained on a labeled dataset with training tokens matched to corresponding vendors.
Yeh discloses: wherein the machine learning model is a neural network (fig.1, fig.2 contemplates use of neural networks in transaction categorization setting), the neural network being trained on a labeled dataset with training tokens matched to corresponding vendors (§IV Datasets contemplates a labeled dataset with comprising vendor category matched to transaction tokens).
It would have been obvious before the effective filing date to one of ordinary skill in the art to modify the method of Range by incorporating the neural network and merchant transaction data setting of Yeh. Both concern the art of token categorization, and the incorporation would have, according to Yeh, solved the problem of accurately categorizing vendors, such as for fraud management (§I).
For claim 2, Range modified by Yeh discloses the method of claim1, as described above. Range modified by Yeh further discloses: the tokenizing the received transaction data further comprising:
receiving the transaction data comprising a transaction description (Range fig.4:401, 0026; Yeh §VI Datasets); and
tokenizing the transaction description to generate the plurality of tokens (ibid).
For claim 3, Range modified by Yeh discloses the method of claim 1, as described above. Range modified by Yeh further discloses: the tokenizing the received transaction data further comprising:
receiving the transaction data comprising a transaction description (Range fig.4:401, 0026, Yeh §IV Datasets); and
tokenizing the transaction description based on spaces within the transaction description to generate the plurality of tokens (fig.3 contemplates tokenizing of space-separated words).
For claim 4, Range modified by Yeh discloses the method of claim 1, as described above. Range modified by Yeh further discloses: the providing the transaction data for downstream processing comprises:
providing the transaction data to an artificial intelligence pipeline (Range fig.1, fig.2 shows overview of artificial intelligence pipeline, such as using artificial intelligence techniques to generate further data including total text attribute predictability).
For claim 6, Range modified by Yeh discloses the method of claim 1, as described above. Range modified by Yeh further discloses: assigning a predictability score of zero to a token that does not have a corresponding dictionary entry (fig.5, 0041: zero or null characters are assigned to tokens without dictionary entries; fig.6: zeros are assigned to tokens not present in the dictionary via a bit mapping1).
For claim 7, Range modified by Yeh discloses the method of claim 1, as described above. Range modified by Yeh further discloses: the training of the neural network comprising:
generating a token-vendor matrix from the labeled dataset, the token-vendor matrix indicating counts of matches between tokens and corresponding vendors (Range fig.1:110, 120 shows generation of token matrix for data processing, with Yeh §IV contemplating application to merchant categories, hence, use of merchant data is tokens at Targ in the matrix);
generating an updated token-vendor matrix by removing tokens with counts lower than a predetermined count threshold (Ragne 0038-39 contemplates removing tokens below a threshold count, hence, original matrix is updated for further processing); and
selecting the training tokens from the updated token-vendor matrix (ibid: training and inference is performed only on remaining tokens).
For claim 8, Range modified by Yeh discloses the method of claim 7, as described above. Range modified by Yeh further discloses: the selecting the training tokens comprising:
normalizing the counts of matches in the updated token-vendor matrix (0038-39: pruning based on ranking constitutes normalizing counts based on relative measures); and
selecting the training tokens and corresponding normalized counts of matches in the updated token-vendor matrix (ibid: the updated tokens are selected for further processing).
For claim 9, Range modified by Yeh discloses the method of claim 7, as described above. Range modified by Yeh further discloses: the selecting the training tokens comprising:
normalizing the counts of matches in the updated token-vendor matrix (0038-39: pruning based on ranking constitutes normalizing counts based on relative measures);
sorting the normalized counts of matches (ibid: the normalized counts are sorted for future processing);
removing the normalized counts of matches below a predetermined normalized counts of matches threshold (ibid: the counts of lower matches and respective tokens are removed for further processing); and
selecting the training tokens and corresponding normalized counts above the predetermined normalized counts of matches threshold (ibid: the updated tokens are selected for further processing).
Claims 11-14, 16-19 recite systems analogous to the above methods and are hence rejected for the same reasons. Furthermore, Range discloses: a non-transitory storage medium storing computer program instructions (fig.1, 9020, 0065); and
a processor configured to execute the computer program instructions to cause operations (fig.10:9010).
Claim(s) 5, 10, 15, 20 are rejected under 35 U.S.C. 103 as being unpatentable over Range (US 20200065710 A1) in view of Yeh ("Merchant category identification using credit card transactions", published 2020) in view of Soni ("TextConvoNet: A convolutional neural network based architecture for text classification", published 2023).
For claim 5, Range modified by Yeh discloses the method of claim 1, as described above. Range further discloses: the retrieving the corresponding dictionary entries comprising:
retrieving the corresponding dictionary entries of the predictability scores generated by a trained neural network (Range fig.2:230-250 contemplates retrieving dictionary entries generated in 220, with Yeh fig.1-2 contemplating use of a neural network to generate entries).
Range modified by Yeh does not disclose: wherein the neural network is convolutional.
Soni discloses: wherein the neural network is convolutional (abstract, p.14250 c.1 ¶2-3, fig.1).
It would have been obvious before the effective filing date to one of ordinary skill in the art to modify the method of Range modified by Yeh by incorporating the CNN technique of Soni. Both concern the art of token classification, and the incorporation would have, according to Soni, allow recognition of patterns in text via an efficient method (p.14250 c.1 ¶2).
For claim 10, Range modified by Yeh discloses the method of claim 1, as described above. . Range modified by Yeh does not disclose: retraining the neural network with a new labeled dataset.
However, the use of multiple datasets for a neural network, such as to solve different problems or to evaluate the neural network architecture performance, is known in the art. In particular, Soni discloses: retraining the neural network with a new labeled dataset (§4.1 discloses the use of various datasets for evaluation).
It would have been obvious before the effective filing date to one of ordinary skill in the art to modify the method of Range modified by Yeh by incorporating the retraining technique Soni. Both concern the art of token classification, and the incorporation would have, according to Soni, allow evaluation on different datasets for performance assessment (§4.1 ¶1).
Claims 15, 20 recite systems analogous to the above methods and are hence rejected for the same reasons.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Bigaj (US 20220309384 A1) discloses feature selection for training machine learning models for bank transaction records.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LIANG LI whose telephone number is (303)297-4263. The examiner can normally be reached Mon-Fri 9-12p, 3-11p MT (11-2p, 5-1a ET).
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. The examiner is available for interviews Mon-Fri 6-11a, 2-7p MT (8-1p, 4-9p ET).
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor Jennifer Welch can be reached on (571)272-7212. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from Patent Center and the Private Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from Patent Center or Private PAIR. Status information for unpublished applications is available through Patent Center or Private PAIR to authorized users only. Should you have questions about access to Patent Center or the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
/LIANG LI/
Primary examiner AU 2143
Prosecution Insights