# US 2023/0274084 A1 · Facilitating generation of fillable document templates > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2023/0274084 A1 - Document ID: US20230274084A1 - Status: Published application - Kind: A1 - Application number: 17/682,911 - Filed: 2022-02-28 - Published: 2023-08-31 - Earliest claimed: 2022-02-28 - Assignees: Adobe Inc. - Applicants: Adobe Inc. - Inventors: Natwar Modani; Muskan Agarwal; Vishesh Kaushik; Aparna Garimella; Akhash N A; Garvit Bhardwaj; Manoj Kilaru; Priyanshu Agarwal - Web page: https://patentgrep.com/patent/US20230274084A1 ## CPC - G06F 40/174: Form filling; Merging - G06F 40/186: Templates - G06F 40/103: Formatting, i.e. changing of presentation of documents (automatic justification G06F40/189; automatic line break hyphenation G06F40/191) - G06F 40/284: Lexical analysis, e.g. tokenisation or collocates - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract Methods and systems are provided for facilitating generation of fillable document templates. In embodiments, a document having a plurality of tokens is obtained. Using a machine learned model, a token state is identified for each token of the plurality of tokens. Each token state indicates whether a corresponding token is a static token that is to be included in a fillable document template or a dynamic token that is to be excluded in the fillable document template. Thereafter, a fillable document template corresponding with the document is generated, wherein for each dynamic token of the document, the fillable document template includes a fillable field corresponding to the respective dynamic token. ## Claims ### Claim 1 A method for converting documents to fillable document templates, the method comprising: obtaining, by a processing device, a document having a plurality of tokens in corresponding regions in the document, wherein the regions include text; identifying, via a machine learned model, a token state for each token of the plurality of tokens, wherein each token state indicates whether a corresponding token is a static token or a dynamic token; and generating a fillable document template corresponding with the document, wherein the fillable document template is generated by, for each dynamic token of the document, removing the dynamic token and replacing the dynamic token with a fillable field enabling input to be provided into the fillable field. ### Claim 2 (depends on 1) The method of claim 1, wherein the generated fillable document template includes the static tokens, or representations thereof. ### Claim 3 (depends on 1) The method of claim 1 further comprising: identifying an entity associated with a first dynamic token; identifying a first static token associated with the entity; and converting the first static token to a second dynamic token based on the first static token and the first dynamic token being associated with the entity. ### Claim 4 (depends on 1) The method of claim 1, wherein the fillable field comprises an unfilled region configured to accept text input. ### Claim 5 (depends on 1) The method of claim 1 further comprising determining a data type associated with at least one dynamic token. ### Claim 6 (depends on 1) The method of claim 1 further comprising providing the fillable document template for display, wherein an indication of a data type or an example of the data type associated with a first dynamic token is presented with the fillable document template. ### Claim 7 (depends on 1) The method of claim 1, wherein the machine learned model is trained using a training dataset including a set of training documents and corresponding token state indicators, the token state indicators determined by grouping similar training documents and identifying text differences between the similar training documents. ### Claim 8 One or more non-transitory computer-readable media having a plurality of executable instructions embodied thereon, which, when executed by one or more processors, cause the one or more processors to perform a method for converting documents to fillable documents, the method comprising: obtaining a set of training documents; identifying a group of similar training documents from the set of training documents; identifying text differences within the group of similar training documents; based on the identified text differences within the group of similar training documents, assigning a dynamic token state indicator to words in the training documents associated with the text differences; and training a machine learning model using at least a subset of the group of similar training documents and corresponding dynamic token state indicators to generate a trained machine learning model that predicts token states for text in new documents. ### Claim 9 (depends on 8) The one or more non-transitory computer-readable media of claim 8, further comprising assigning a static token state indicator to words not associated with the text differences, wherein the machine learning model is trained using the corresponding dynamic token state indicators and the corresponding static token state indicators. ### Claim 10 (depends on 8) The one or more non-transitory computer-readable media of claim 8, wherein the group of similar training documents is identified using a bag-of-words similarity analysis. ### Claim 11 (depends on 8) The one or more non-transitory computer-readable media of claim 8, wherein the group of similar training documents is identified using a bag-of-words similarity analysis and a word-order similarity analysis. ### Claim 12 (depends on 11) The one or more non-transitory computer-readable media of claim 11, wherein the bag-of-words similarity analysis includes using a MinHash-local sensitive hashing scheme. ### Claim 13 (depends on 11) The one or more non-transitory computer-readable media of claim 11, wherein the word-order similarity analysis includes using a diff-based similarity analysis on a pairwise basis. ### Claim 14 (depends on 8) The one or more non-transitory computer-readable media of claim 8, wherein identifying text differences within the group of similar training documents comprises using a diff-match-patch set of algorithms to analyze pairs of training documents. ### Claim 15 (depends on 8) The one or more non-transitory computer-readable media of claim 8 further comprising: performing named entity recognition or noun phrase detection; for an identified entity or noun phrase that corresponds with a word assigned a dynamic token state indicator, assigning the dynamic token state indicator to other words associated with the identified entity or noun phrase. ### Claim 16 (depends on 8) The one or more non-transitory computer-readable media of claim 8, wherein training the machine learning model further comprises using at least a portion of training documents in a second group of similar training documents. ### Claim 17 (depends on 8) The one or more non-transitory computer-readable media of claim 8, wherein the machine learning model is used to predict token state indicators for tokens associated with a new document. ### Claim 18 A system comprising: one or more processors; and one or more memory coupled with the one or more processors, the one or more processors to perform operations comprising: identifying a first document as similar to a second document; identifying text differences between the first document and the second document identified as similar to the first document; for the first document, assigning token state indicators for tokens in the first document based on the identified text differences between the first document and the second document, the token state indicators indicating whether a corresponding token is a static token or a dynamic token; and training, using the tokens in the first document and the corresponding token state indicators, a machine learning model to predict token states for text in new documents. ### Claim 19 (depends on 18) The system of claim 18, wherein the computing system is further caused to identify the first document and the second document are similar based on a bag-of-words similarity analysis and a word-order similarity analysis. ### Claim 20 (depends on 18) The system of claim 18, wherein a token state indicator indicates a token is a dynamic token when a text difference between the first document and the second document is identified in association with the token.