DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim 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, 2, 8-12 and 18-20 is/are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Zhong (“Image-Based Table Recognition: Data, Model, and Evaluation”, Computer Vision – ECCV 2020: 16th European Conference, Glasgow, UK, August 23-28, 2020, Proceedings, Part XXI, 1 January 2020, Springer International Publishing, XP093215701, ISSN: 0302-9743, ISBN: 978-3-030-58589-1, vol. 12366, pages 564-580).
-Regarding claim 1, Zhong discloses a data processing method comprising: obtaining a to-be-processed table image (input table images, section 4); determining a table recognition result based on the table image and a generative table recognition policy (when the structure decoder recognizes a new cell, the cell decoder is triggered and sues the hidden state of the structure decoder to compute the attention for recognizing the content of the new cell, section 4), wherein the generative table recognition policy indicates that the table recognition result of the table image is to be determined by using a markup language (i.e., HTML and HTML tags, section 4) and a non-overlapping attribute of a bounding box (i.e., individual cell that is detected as part of the table structure, section 4), wherein the bounding box indicates a position of a text included in a cell in a table associated with the table image (recognized cell’s position and content of the new cell, section 4), and the table recognition result indicates a global structure and content that are included in the table (the structure decoder only generates the HTML tags that define the structure of the table, section 4); and outputting the table recognition result (the outputs of the two decoders can be easily merged to get the final HTML representation of the table, section 4); wherein determining the table recognition result based on the table image and the generative table recognition policy comprises: obtaining the table recognition result through iteration processing based on a table image feature and the markup language (RNN, section 4).
-Regarding claim 2, Zhong further discloses the non-overlapping attribute of the bounding box indicates that areas corresponding to all cells comprised in the table do not overlap (cells are separate and contain separate content, section 4).
-Regarding claim 8, Zhong further discloses correcting the table recognition result based on the table image; and outputting a corrected table recognition result (Tree-Edit-Distance-Based Similarity, section 5).
-Regarding claim 9, Zhong further discloses performing feature extraction on the table image to obtain the table image feature (the encoder is a convolutional neural network (CNN) that captures the visual features of input table images, section 4).
-Regarding claim 10, Zhong further discloses the table recognition result is identified by using a hypertext markup language (HTML), an extensible markup language (XML), or LaTex (HTML, XML, see abstract).
-Regarding claim 11, Zhong discloses a data processing chip comprising: a logic circuit configured to perform operations including: obtaining a to-be-processed table image (input table images, section 4); determining a table recognition result based on the table image and a generative table recognition policy (when the structure decoder recognizes a new cell, the cell decoder is triggered and sues the hidden state of the structure decoder to compute the attention for recognizing the content of the new cell, section 4), wherein the generative table recognition policy indicates that the table recognition result of the table image is to be determined by using a markup language (i.e., HTML and HTML tags, section 4) and a non-overlapping attribute of a bounding box (i.e., individual cell that is detected as part of the table structure, section 4), wherein the bounding box indicates a position of a text included in a cell in a table associated with the table image (recognized cell’s position and content of the new cell, section 4), and the table recognition result indicates a global structure and content that are included in the table (the structure decoder only generates the HTML tags that define the structure of the table, section 4); and outputting the table recognition result (the outputs of the two decoders can be easily merged to get the final HTML representation of the table, section 4); wherein of determining the table recognition result based on the table image and the generative table recognition policy comprises: obtaining the table recognition result through iteration processing based on a table image feature and the markup language (RNN, section 4).
-Regarding claim 12, Zhong further discloses the non-overlapping attribute of the bounding box indicates that areas corresponding to all cells comprised in the table do not overlap (cells are separate and contain separate content, section 4).
-Regarding claim 18, Zhong further discloses the logic circuit is further configured to perform an operations of: correcting the table recognition result based on the table image (Tree-Edit-Distance-Based Similarity, section 5); and outputting a corrected table recognition result (output, section 5).
-Regarding claim 19, Zhong further discloses the logic circuit is further configured to perform an operation of: performing feature extraction on the table image to obtain the table image feature (the encoder is a convolutional neural network (CNN) that captures the visual features of input table images, section 4).
-Regarding claim 20, Zhong discloses a data processing system comprising: a memory storing executable instructions (RAM, section 6.1); a processor configured to execute the executable instructions to perform operations (GPU, section 6.1) including: obtaining a to-be-processed table image (input table images, section 4); determining a table recognition result based on the table image and a generative table recognition policy (when the structure decoder recognizes a new cell, the cell decoder is triggered and sues the hidden state of the structure decoder to compute the attention for recognizing the content of the new cell, section 4), wherein the generative table recognition policy indicates that the table recognition result of the table image is to be determined by using a markup language (i.e., HTML and HTML tags, section 4) and a non-overlapping attribute of a bounding box (i.e., individual cell that is detected as part of the table structure, section 4), wherein the bounding box indicates a position of a text included in a cell in a table associated with the table image (recognized cell’s position and content of the new cell, section 4), and the table recognition result indicates a global structure and content that are included in the table (the structure decoder only generates the HTML tags that define the structure of the table, section 4); and outputting the table recognition result (the outputs of the two decoders can be easily merged to get the final HTML representation of the table, section 4); wherein the operation of determining the table recognition result based on the table image and the generative table recognition policy comprises: obtaining the table recognition result through iteration processing based on a table image feature and the markup language (RNN, section 4)
Allowable Subject Matter
Claims 4-7 and 14-17 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.
Response to Arguments
Applicant's arguments filed 7/21/2026 have been fully considered but they are not persuasive.
Applicant argues that Zhong's EDD model "does not use a bounding box at all in its recognition pipeline." Applicant further argues that Zhong's references to bounding boxes relate only to table detection and dataset generation.
The Examiner respectfully disagrees. The claim does not recite any particular recognition architecture. It recites that the generative table recognition policy "indicates that the table recognition result … is to be determined by using a markup language and a non-overlapping attribute of a bounding box." Based on the broadest reasonable interpretation, this reads on any policy in which the result is determined using markup together with positional regions of cell text that do not overlap one another. Zhong teaches this in the following ways:
Cell-localized regions. When the structure decoder generates a new cell, it focuses its attention around that cell. Its hidden state is passed to the cell decoder so that the cell decoder attends to the corresponding cell in the image (Sec. II.B; Sec. IV). The cell decoder's aggregated attention for each cell is focused on that cell's own content (Fig. 7(d)).
Non-overlapping. Zhong ensures a one-to-one match between the cells generated by the structure decoder and the sequences generated by the cell decoder (Sec. IV). Each cell's content is localized and recognized separately from every other cell's content, which is a non-overlapping relationship between cell positions.
Cell location with markup. Zhong expressly discloses adding cell coordinates and a branch of the EDD network that also predicts cell location, so that cell location is used together with the HTML output in recognizing cell content (Sec. VII).
Bounding boxes and text. Zhong uses a bounding box to extract the text within it and compare that text with the HTML annotation (Sec. III). This shows that Zhong uses bounding boxes as indicators of text position in connection with the markup representation.
Applicant further argues that Zhong "is completely silent with regard to an iterative approach" and "does not mention the term iteration."
The Examiner respectfully disagrees. Zhong teaches iteration processing as claimed:
Table image feature. Zhong's encoder is a CNN that captures the visual features of the input table image (Zhong, Fig. 1 caption; Sec. IV). This feature map is a table image feature.
Markup language. The structure decoder generates HTML tags that define the structure of the table (Sec. IV; Fig. 2). HTML is a markup language, and the final result is an HTML representation of the table (Fig. 1, "Merged output").
Iteration processing. The structure decoder and the cell decoder are recurrent neural networks, implemented as single-layer LSTMs (Sec. IV; Sec. VI.A). As shown in Fig. 1, each recurrent unit R_s receives the hidden state and output token from the previous step. It also receives an attention-weighted view of the encoder feature map from attention network A_s. From these it produces the next token, and the process repeats until the table is complete. The cell decoder (R_c, A_c) works the same way, one character per step, for each cell. Each step builds on the result of the prior step, so the recognition result is obtained through repeated, successive processing. That is iteration processing under any reasonable interpretation of the term.
Based on both. Zhong confirms that each step depends on both the image feature and the previously generated markup. The structure decoder attends to the image region of the cell being generated. It predicts the closing </td> tag by relying on the language model, meaning the previously generated markup tokens, rather than on the feature map (Sec. VI.C; Fig. 7(a)–(c)). Zhong also samples the output with beam search at inference (Sec. VI.A), which extends partial markup sequences step by step.
Accordingly, Zhong teaches obtaining the table recognition result through iteration processing based on a table image feature and the markup language.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PING Y HSIEH whose telephone number is (571)270-3011. The examiner can normally be reached Monday-Friday, 9am-4pm.
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, Jennifer Mehmood can be reached at (571) 272-2976. 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.
/PING Y HSIEH/Primary Examiner, Art Unit 2664
