DETAILED ACTION
Claims 1-20 are pending in the Instant Application.
Claims 1-4, 6-8, 10-16, 18--20 are rejected (Non-Final Rejection).
Claims 5, 9 and 17 are objected to.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Priority
The Instant Application, filed 02/13/2025, is a continuation-in-part of application 18/545.020, filed 12/19/2023, which was a continuation-in-part of application 18/238,135, filed 08/25/2023. Therefore, the earliest effective filing date of the claims is 08/25/2023, 12/19/2023 or 02/13/2025 depending in what application the limitations were recited in.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-4, 6, 7, 10-15 and 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Xian et al. (“Xian”), United States Patent Application Publication No. 2021/0232908, in view of PRASAD et al. (“Prasad”), United States Patent Application Publication No. 2021/0019287, in further view of Saadi, United States Patent Application Publication No. 2023/0297831.
As per claim 1, Xian discloses a computer-implemented method comprising:
receiving input data that is organized into a set of rows and a set of columns ([0049]-[0050] wherein data set is input that includes tables, wherein tables are organized in a set of rows and columns (See [Fig. 2] wherein the data is a set of rows and a set of columns));
maintaining a machine learning column model that is trained on the tabular data ([0089] wherein the machine learning column model (cell neural network encoder model in the prior art) is maintained and trained using tabular data);
supplying header row identification data and features of the input data to the machine learning column model to generate column label data that applies a set of defined labels to the set of columns ([0103] wherein header row identification data and features of the input data (recognized as the encoded column in the prior art) are supplied to generate a column label (recognized as the schema labels in the prior art)), wherein: for each column of the set of columns:
predicting one of the set of defined labels for the column ([0099] wherein a label is predicted);
determining a column confidence score from the machine learning column model for a column label prediction ([0099] wherein the similarity scores captures the confidence of the label, making it a confidence score);
and generating output data that is organized into rows and columns, wherein columns of the output data are labelled based on the column label data( [0106] wherein a graphical interface can provide the dataset including the column labels determined), but does not disclose maintaining a machine learning header model that is trained on tabular data with header rows; supplying the input data as input to the machine learning header model to generate header row identification data that identifies a set of header rows that is a subset of the set of rows; generating a first notification seeking first user feedback when the column confidence score is below a first threshold for the column; receiving the first user feedback; and applying the first user feedback to predict the one of the set of defined labels for the column. However, Prasad teaches maintaining a machine learning header model that is trained on tabular data with header rows ([0045] wherein a model is trained using tabular data for each content object, wherein the header content object as described in [0039]); supplying the input data as input to the machine learning header model to generate header row identification data that identifies a set of header rows that is a subset of the set of rows ([0047] wherein the input data is supplied to a machine learning model to generate the types of content objects, including the header content object (see [0048]), and based on the content object identification of the model, header row identification data is identified along with a set of header rows), but does not teach generating a first notification seeking first user feedback when the column confidence score is below a first threshold for the column; receiving the first user feedback; and applying the first user feedback to predict the one of the set of defined labels for the column. However, Saadi teaches generating a first notification seeking first user feedback when the label confidence score is below a first threshold for the column ([0063] wherein if the confidence score is too low, a user may be provided with a selection of labels for the item); receiving the first user feedback ([0063] wherein the user feedback is received); and applying the first user feedback to predict the one of the set of defined labels ([0059] wherein the feedback data is used to retrain the model) .
Xian determines header row data, but does not expressly use a model to determine the header row. Prasad describes using a model to determine the header information and generates header row data as does the claim. One could use the model in Prasad instead of the rule based system in Xian to teach the claimed invention. It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to combine the method of determine column labels using a machine learning model in Xian, with the column label row being determined by a machine learning model as in Prasad in order to be able to use the same model for a variety of formatted data and use machine learning to identify the column header data instead of a human. Xian labels column data but does expressly use user feedback when the confidence score for the label is low. One could apply the user feedback in Saadi, with the column labels in Xian to teach the claimed invention. It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to combine the method of labeling column data using a trained machine learning model in Xian with user feedback if the confidence in labels is low as in Saadi in order to get new training data to constantly improve the model.
As per claim 2, note the rejection of claim 1 where Xian, Prasad and Saadi are combined. The combination teaches the method of claim 1. Xian further discloses wherein, to each of the set of columns, the column label data applies one of the set of defined labels ([0067] wherein the labels are those used to train the model).
As per claim 3, note the rejection of claim 1 where Xian, Prasad and Saadi are combined. The combination teaches the method of claim 1. Xian further discloses the column label prediction (see claim 1). Saadi further teaches generating a user interface configured to display the first notification and generate one or more user activatable components allowing a user to validate or override the label prediction ([0063] and [Fig.8] wherein the user is provided a user activatable component where the user can validate the label by selecting “Onions” or override the label by selecting “Mushrooms.” )
As per claim 4, note the rejection of claim 1 where Xian, Prasad and Saadi are combined. The combination teaches the method of claim 1. Saadi further teaches wherein the first notification comprises sending an email to a user, generating a pop-up menu on a user interface, or sending a text message (Examiner notes the use of “or” where only one option is selected [Fig. 8] wherein a pop-up menu is shown with different options).
As per claim 6, note the rejection of claim 1 where Xian, Prasad and Saadi are combined. The combination teaches the method of claim 1. Prasad further teaches wherein: supplying the header row identification data and features of the input data to the machine learning column model includes, for each column of the set of columns, supplying features of the column to each of the plurality of column- specific machine learning models; and for each column of the set of columns, the features include data values from the column ([0045] wherein the model is trained based on column data and properties of that content type).
As per claim 7, note the rejection of claim 1 where Xian, Prasad and Saadi are combined. The combination teaches the method of claim 1. Prasad further teaches using the machine learning header model to determine a header row score for a subset of the set of rows; and in response to the header row score for a row exceeding the first threshold, including the row in the header row identification data ([0047] wherein the model is used to determine the headers based on header content objects, and from there the header row score is determined to identify the header by threshold as described in [0080]), but does not disclose generating a second notification seeking second user feedback when the header row score for the row is below the first threshold; receiving the second user feedback; and applying the second user feedback to train the machine learning header model. However, Saadi teaches a notification seeking user feedback when the score for the identification is below the first threshold ([0063] wherein if the confidence score is too low, a user may be provided with a selection of identifications for the item); receiving the user feedback [0063] wherein the user feedback is received);; and applying the user feedback to train the machine learning model ([0059] wherein the feedback data is used to retrain the model) ..
Prasad determines header rows, and scores those, but does not expressly use user feedback when the confidence score for the label is low. One could apply the user feedback in Saadi, with the column type in Prasad to teach the claimed invention. It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to combine the method of labeling columns using a trained machine learning model in Prasad with user feedback if the confidence in labels is low as in Saadi in order to get new training data to constantly improve the model.
As per claim 10, note the rejection of claim 1 where Xian, Prasad and Saadi are combined. The combination teaches the method of claim 1. Xian discloses column labels, but does not expressly describe those labels triggering feedback. Saadi further teaches the machine learning column model backpropagating the label prediction to a first predicted column of the set of columns after receiving the first user feedback ([0063] wherein the user can choose onions and backpropagate the column label).
As per claim 11, note the rejection of claim 1 where Xian, Prasad and Saadi are combined. The combination teaches the method of claim 1. Xian teaches column labels, but does not expressly describe those labels triggering feedback. Saadi further teaches the machine learning column model backpropagating the label prediction to a predicted label of the set of labels that had the confidence score lower than the first threshold after receiving the first user feedback ([0063] wherein the user can choose onions and backpropagate the column label, after choosing the label with a score lower than the threshold).
As per claim 12, Xian discloses a system comprising: processor hardware ([0128]); and memory hardware configured to store instructions ([0141]) that, when executed by the processor hardware, cause the processor hardware to perform the method of claim 1, taught by the combination of Xian, Prasad and Saadi. Thus, claim 12 is rejected for the same rationale and reasoning as claim 1 above.
As per claim 13, claim 13 is the system that performs the method of claim 3 and is rejected for the same rationale and reasoning.
As per claim 14, claim 14 is the system that performs the method of claim 4 and is rejected for the same rationale and reasoning.
As per claim 15, claim 15 is the system that performs the method of claim 7 and is rejected for the same rationale and reasoning.
As per claim 18, claim 18 is the system that performs the method of claim 10 and is rejected for the same rationale and reasoning.
As per claim 19, claim 19 is the system that performs the method of claim 11 and is rejected for the same rationale and reasoning.
As per claim 20, claim 20 is a computer program product that includes instructions to perform the method of claim 1 and is rejected for the same rationale and reasoning.
Claims 8 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Xian in view Prasad, in further view of Saadi, in further view of Smart et al. (“Smart”), United States Patent Application Publication No. 2021/0334275.
As per claim 8, note the rejection of claim 1 where Xian, Prasad and Saadi are combined. The combination teaches the method of claim 1, but does not does disclose maintaining a machine learning junk model that is trained on tabular data; supplying features of the input data to the machine learning junk model to generate junk scores for the set of rows; and for any row of the set of rows having a junk score above the first threshold, excluding the row from the output data. However, Smart teaches maintaining a machine learning junk model that is trained on tabular data; supplying features of the input data to the machine learning junk model to generate junk scores for the set of rows; and for any row of the set of rows having a junk score above the first threshold, excluding the row from the output data ([0004] wherein the predictive machine learning model determines a score for the row that is “junk” in that is not associated with another row, and the row is excluded if the score is above the threshold level of match).
Xian, Prasad and Smart describe analyzing data sets. One could include the junk removal in Smart with the table data extraction of Xian and Prasad to teach the claimed invention. It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to combine the method of labeling data using a model to determine header rows in the combination of Xian and Prasad with the removal of rows with junk data in Smart in order to avoid displaying rows that do not apply.
As per claim 16, claim 16 is the system that performs the method of claim 8 and is rejected for the same rationale and reasoning.
Allowable Subject Matter
Claims 5, 9 and 17 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
The following is a statement of reasons for the indication of allowable subject matter:
As per claim 5, the following claim language in claim 5 including, “determining a highest scoring model of the plurality of column- specific machine learning models; selectively applying the one of the set of defined labels corresponding to the highest scoring model to the column; and applying the undefined label to the column in response to the score of the highest scoring model being less than the first threshold,“ are neither anticipated nor obvious over the prior art on record. Thus, the claim is objected to.
As per claim 9, the following language in claim 9 including, “generating a third notification seeking third user feedback when the junk score is below the first threshold and exceeds a second threshold; receiving the third user feedback; and applying the third user feedback to train the machine learning junk model,” are neither anticipated nor obvious over the prior art on record. Thus, the claim is objected to. Claim 17 is the system that performs the method of claim 9 and is objected to for the same rationale and reasoning.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KANNAN SHANMUGASUNDARAM whose telephone number is (571)270-7763. The examiner can normally be reached M-F 9:00 AM -6:00 PM.
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, Charles Rones can be reached at (571) 272-4085. 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.
/KANNAN SHANMUGASUNDARAM/Primary Examiner, Art Unit 2168