Method and system for determining structural blocks of a document
This disclosure relates generally to document processing, and more particularly to method and system for determining structural blocks of a document. In one embodiment, the method may include extracting text from the document, the text including text lines. The method may further include generating a feature vector for each of the text lines, the feature vector for the text line including a set of feature values for a set of corresponding features in the text line. The method may further include creating an input matrix for each of the text lines, the input matrix for the text line including a set of feature vectors corresponding to a set of neighboring text lines along with the text line. The method may further include determining a structural block tag for each of the text lines based on the corresponding input matrix using a machine learning model.
1. A method of determining document structural blocks, the method implemented by one or more document analysis devices and comprising:
extracting text from a document, wherein the text comprises a plurality of text lines;
generating a feature vector for each of the text lines, wherein the feature vectors each comprise a set of feature values for a set of corresponding features in one of the text lines;
creating an input matrix for each of the text lines, wherein the input matrices each comprise a set of feature vectors corresponding to a set of neighboring ones of the text lines along with one of the text lines; and
determining a structural block tag for each of the text lines based on the corresponding one of the input matrices using a machine learning model.
2. The method of claim 1 , further comprising:
receiving an image document;
performing an optical text recognition on the image document to generate the document; and
applying a text extraction tool with a pre-defined or a dynamic threshold on the document to extract the text lines from the document.
3. The method of claim 1 , wherein at least one of:
the set of corresponding features comprise at least one of a positional feature, a font feature, a count feature, a spacing feature, or a semantic feature; or
the set of feature values comprises at least one of one or more positional coordinates, a font size, a font weight, a font relative weight, one or more flags for one or more font styles, or a length.
4. The method of claim 1 , wherein the set of neighboring text lines comprises a pre-defined set of preceding text lines and a pre-defined set of successive text lines.
5. The method of claim 1 , wherein the structural block tag comprises a paragraph tag or a non-paragraph tag, wherein the paragraph tag comprises one of a footnote tag, a paragraph start tag, a single line tag, or a paragraph tag, and wherein the non-paragraph tag comprises one of a section header tag, a list tag, a table of content tag, a title tag, or an un-classified tag.
6. The method of claim 1 , wherein the machine learning model is a sequence to sequence machine learning model and the method further comprises training the machine learning model with training data, wherein the training data comprises a set of texts extracted from a set of documents, and, for each of the set of texts, a set of text lines with a corresponding set of feature vectors and a corresponding set of pre-defined structural block tags.
7. A document analysis device, comprising a non-transitory computer readable medium having programmed instructions stored thereon and a processor configured for being communicatively coupled with the non-transitory computer readable medium and further configured to execute the stored programmed instructions to:
extract text from a document, wherein the text comprises a plurality of text lines;
generate a feature vector for each of the text lines, wherein the feature vectors each comprise a set of feature values for a set of corresponding features in one of the text lines;
create an input matrix for each of the text lines, wherein the input matrices each comprise a set of feature vectors corresponding to a set of neighboring ones of the text lines along with one of the text lines; and
determine a structural block tag for each of the text lines based on the corresponding one of the input matrices using a machine learning model.
8. The document analysis device of claim 7 , wherein the processor is further configured to execute the stored programmed instructions to:
receive an image document;
perform an optical text recognition on the image document to generate the document; and
apply a text extraction tool with a pre-defined or a dynamic threshold on the document to extract the text lines from the document.
9. The document analysis device of claim 7 , wherein at least one of:
the set of corresponding features comprise at least one of a positional feature, a font feature, a count feature, a spacing feature, or a semantic feature; or
the set of feature values comprises at least one of one or more positional coordinates, a font size, a font weight, a font relative weight, one or more flags for one or more font styles, or a length.
10. The document analysis device of claim 7 , wherein the set of neighboring text lines comprises a pre-defined set of preceding text lines and a pre-defined set of successive text lines.
11. The document analysis device of claim 7 , wherein the structural block tag comprises a paragraph tag or a non-paragraph tag, wherein the paragraph tag comprises one of a footnote tag, a paragraph start tag, a single line tag, or a paragraph tag, and wherein the non-paragraph tag comprises one of a section header tag, a list tag, a table of content tag, a title tag, or an un-classified tag.
12. The document analysis device of claim 7 , wherein the machine learning model is a sequence to sequence machine learning model and the processor is further configured to execute the stored programmed instructions to train the machine learning model with training data, wherein the training data comprises a set of texts extracted from a set of documents, and, for each of the set of texts, a set of text lines with a corresponding set of feature vectors and a corresponding set of pre-defined structural block tags.
13. A non-transitory computer readable medium having stored thereon instructions for determining document structural blocks comprising executable code which when executed by one or more processors, causes the processors to:
extract text from a document, wherein the text comprises a plurality of text lines;
generate a feature vector for each of the text lines, wherein the feature vectors each comprise a set of feature values for a set of corresponding features in one of the text lines;
create an input matrix for each of the text lines, wherein the input matrices each comprise a set of feature vectors corresponding to a set of neighboring ones of the text lines along with one of the text lines; and
determine a structural block tag for each of the text lines based on the corresponding one of the input matrices using a machine learning model.
14. The non-transitory computer readable medium of claim 13 , wherein the executable code when executed by the processors further causes the processors to:
receive an image document;
perform an optical text recognition on the image document to generate the document; and
apply a text extraction tool with a pre-defined or a dynamic threshold on the document to extract the text lines from the document.
15. The non-transitory computer readable medium of claim 13 , wherein at least one of:
the set of corresponding features comprise at least one of a positional feature, a font feature, a count feature, a spacing feature, or a semantic feature; or
the set of feature values comprises at least one of one or more positional coordinates, a font size, a font weight, a font relative weight, one or more flags for one or more font styles, or a length.
16. The non-transitory computer readable medium of claim 13 , wherein the set of neighboring text lines comprises a pre-defined set of preceding text lines and a pre-defined set of successive text lines.
17. The non-transitory computer readable medium of claim 13 , wherein the structural block tag comprises a paragraph tag or a non-paragraph tag, wherein the paragraph tag comprises one of a footnote tag, a paragraph start tag, a single line tag, or a paragraph tag, and wherein the non-paragraph tag comprises one of a section header tag, a list tag, a table of content tag, a title tag, or an un-classified tag.
18. The non-transitory computer readable medium of claim 13 , wherein the machine learning model is a sequence to sequence machine learning model and the executable code when executed by the processors further causes the processors to train the machine learning model with training data, wherein the training data comprises a set of texts extracted from a set of documents, and, for each of the set of texts, a set of text lines with a corresponding set of feature vectors and a corresponding set of pre-defined structural block tags.