System and method for document conversion to a template
View Patent ↗Systems and methods are provided for converting a document comprising text to a template using a model, and for improving the model so used. The systems and methods presented comprise a local computing device, a remote computing device, communication modules, information on datatypes and patterns that may be converted from document text to template fields, rules regarding document context, and verification criteria. A user may use the systems and methods to convert a document comprising text into a template comprising fields derived from the document text, and optionally may use the systems and methods to improve the model based on a set or subset of documents. The present invention solves problems with the currently available systems and methods of converting document text to a template.
1. A system for template creation from document text, the system comprising:
at least one user device, which has stored on it a plurality of documents, which plurality of documents comprise document text and document context;
at least one server device, which comprises a plurality of input/output components, a plurality of processors, a plurality of memory components, a plurality of databases, and a plurality of computer-readable instructions stored in non-transitory computer-readable media; and
wherein the server device has stored on it information about a plurality of datatypes and a plurality of patterns, which together comprise a model; and
wherein the server device employs a shallow parsing module, a pattern recognition module, and a verification module.
2. A document conversion method for template creation from document text, the method comprising:
a server device receives a document text from a user device,
the server device identifies a context of the document text,
the server device conducts shallow parsing, wherein shallow parsing comprises:
the server device identifies a plurality of datatypes relevant to the document text and the document context, then
the server device searches the document text for one of a plurality of datatypes, then
the server device finds an instance of each one of the plurality of datatypes in the document text, then
the server device parses the text around the instance of each one of the plurality of datatypes for a text-instance, then
the server device adds the text-instance to a list of text-instances, then
the server device assesses whether all the document text has been searched; and
the server device conducts pattern recognition, wherein pattern recognition comprises:
the server device identifies a plurality of patterns relevant to the document text and the document context, then
the server device searches the document text for one of the plurality of patterns, then
the server device finds a pattern-instance of that one of the plurality of patterns in the document text, then
the server device adds the pattern-instance to a list of pattern-instances, then
the server device assesses whether all the document text has been searched for that one of the plurality of patterns; and
the server device assembles a list of all datatype text-instances and pattern-instances identified in the document text,
the server device sends the list of all datatype text-instances and pattern-instances identified to the user device, whereupon
the user device replaces the identified datatype text-instances and pattern-instances in the document text with a set of tagged fields, and
the user device creates a new template document with the set of tagged fields.
3. The method of claim 2 , wherein the method further comprises the server device determines that all the document text has not been searched for the one of the plurality of datatypes, and thereafter the server device continues searching the document text for the one of the plurality of datatypes.
4. The method of claim 2 , wherein the method further comprises the server device determines that all the document text has been searched for the one of the plurality of datatypes, and thereafter the server device verifies each text-instance with datatype verification criteria for that one of the plurality of datatypes, saves each verified text-instance to a list of verified shallow parsing datatype text-instances, and checks for another one of the plurality of datatypes to search.
5. The method of claim 4 , wherein the method further comprises the server device determining that there is another one of the plurality of datatypes to search, the server device thereafter iterates prior steps of the shallow parsing method to search for a next one of the plurality of datatypes.
6. The method of claim 4 , wherein the method further comprises the server device determining that there is not another one of the plurality of datatypes to search, then the server device stops the shallow parsing method.
7. The method of claim 2 , wherein the method further comprises the server device determines that all the document text has not been searched for that one of the plurality of patterns, and thereafter the server device continues searching the document text for that one of the plurality of patterns.
8. The method of claim 2 , wherein the method further comprises the server device determines that all the document text has been searched for that one of the plurality of patterns, and thereafter the server device verifies each pattern-instance with verification criteria for that one of the plurality of patterns, saves each verified pattern-instance to a list of verified pattern-instances, and checks for another one of the plurality of patterns to search.
9. The method of claim 8 , wherein, if each applicable one of the plurality of patterns has not been searched for, the method further comprises the server device determining that there is another one of the plurality of patterns to search, and the server device thereafter iterates prior steps of the pattern recognition method to search for a next one of the plurality of patterns.
10. The method of claim 8 , wherein, if each applicable one of the plurality of patterns has been searched for, the method further comprises the server device determining that there is not another one of the plurality of patterns to search, and the server device stops the pattern recognition method.
11. The method of claim 2 , wherein the user device stores a plurality of documents, and sends the document text and the set of document context to the server device; thereafter the user device receives from the server device a plurality of verified datatype text-instances and a plurality of verified pattern instances; thereafter the user device modifies the document text with the plurality of verified datatype text-instances and the plurality of verified pattern instances to create the document text with the set of tagged fields; thereafter the user device uses the document text with the set of tagged fields to create a template document.
12. A plurality of computer-readable instructions stored in non-transitory computer-readable media configured to carry out methods for template creation from a document text, as described in claim 2 .