Automatically identifying chunks in sets of documents
Machine learning, artificial intelligence, and other computer-implemented methods are used to identify various semantically important chunks in documents, automatically label them with appropriate datatypes and semantic roles, and use this enhanced information to assist authors and to support downstream processes. Chunk locations, datatypes, and semantic roles can often be automatically determined from what is here called “context”, to wit, the combination of their formatting, structure, and content; those of adjacent or nearby content; overall patterns of occurrence in a document, and similarities of all these things across documents (mainly but not exclusively among documents in the same document set). Similarity is not limited to exact or fuzzy string or property comparisons, but may include similarity of natural language grammatical structure, ML (machine learning) techniques such as measuring similarity of word, chunk, and other embeddings, and the datatypes and semantic roles of previously-identified chunks.
1. A method implemented on a computer system executing instructions for analyzing and annotating documents, the method comprising:
importing documents in a document set;
performing visual extraction of the imported documents, including creating signatures for document parts;
automatically identifying a hierarchical structure of chunks within individual documents in the document set (a) based on the visual extraction, content, and contexts in the individual document; and (b) based on patterns of visual extraction and content across the documents in the document set, wherein the hierarchical structure includes small chunks comprising series of words from within individual sentences;
for at least some of the small chunks, in a process separate from identifying the small chunks, automatically selecting text from sentences surrounding the small chunks as labels for semantic roles played by the small chunks in a transaction described by the individual documents;
standardizing the labels for semantic roles across the documents in the document set; and
annotating documents in the document set wherein the annotations include locations of the identified small chunks and standardized labels for the semantic roles played by the identified small chunks at those locations.
2. The computer-implemented method of claim 1 , further comprising:
assembling the document set by clustering documents into the document set based on similarity of content and/or geometric layout.
3. The computer-implemented method of claim 1 , wherein automatically identifying the hierarchical structure of chunks within individual documents in the document set is further (c) based on identifying counterpart chunks in different documents in the document set, wherein counterpart chunks are different chunks in different documents that play a same semantic role within their respective documents.
4. The computer-implemented method of claim 3 , wherein identifying counterpart chunks in different documents comprises:
identifying content that is different in different documents but occurs within substantially similar contexts within the different documents.
5. The computer-implemented method of claim 3 , wherein identifying counterpart chunks in different documents comprises:
identifying content that is substantially the same in different documents.
6. The computer-implemented method of claim 1 , further comprising:
annotating some of the identified chunks with metadata describing the chunk, wherein identifying counterpart chunks in different documents is based on similarity of the metadata.
7. The computer-implemented method of claim 1 , wherein identifying chunks based on patterns across the documents in the document set comprises:
identifying, in an individual document, a chunk that is commonly occurring in the documents of the document set but does not occur in the individual document.
8. The computer-implemented method of claim 1 , wherein the identified chunks comprise:
field chunks that contain content within the documents suitable for use as fields in document templates; and
structural chunks that contain content comprising structures within a geometric layout of the documents.
9. The computer-implemented method of claim 8 , wherein some of the field chunks are hierarchical and contain other chunks as sub-chunks.
10. The computer-implemented method of claim 1 , wherein some of the identified chunks contain content that is descriptive of semantic roles played by other chunks.
11. The computer-implemented method of claim 1 , wherein the annotations further comprise datatypes of the identified chunks.
12. The computer-implemented method of claim 1 , wherein identifying chunks based on visual extraction comprises:
identifying spatial boundaries of structural chunks using machine learning inference trained on tiles of page images.
13. The computer-implemented method of claim 1 , wherein identifying chunks based on visual extraction comprises:
identifying spatial boundaries of structural chunks using artificial intelligence-based visual recognition of geometric patterns of the visual extraction.
14. The computer-implemented method of claim 1 , wherein identifying chunks based on visual extraction comprises:
identifying structural chunks based on visual extraction of non-text structural features, wherein the non-text structural features comprise at least one of a figure, a table, a sidebar, a footnote, and a page header or footer.
15. The computer-implemented method of claim 1 , wherein identifying chunks based on content comprises:
identifying chunks using AI techniques for topic estimation.
16. The computer-implemented method of claim 1 , wherein identifying chunks based on content comprises:
using few-shot Named Entity recognition techniques to identify chunks within the set of documents.
17. The computer-implemented method of claim 1 , further comprising:
receiving user corrections for incorrectly identified chunks; and
improving the step of automatically identifying the hierarchical structure of chunks in response to the user corrections.
18. A non-transitory computer-readable storage medium storing executable computer program instructions for analyzing and improving documents, the instructions executable by a computer system and causing the computer system to perform a method comprising:
importing a document set that contains a plurality of documents;
performing visual extraction of the imported documents, including creating signatures for document parts;
automatically identifying a hierarchical structure of chunks within individual documents in the document set (a) based on the visual extraction, content, and contexts in the individual document; and (b) based on patterns of visual extraction and content across the documents in the document set, wherein the hierarchical structure includes small chunks comprising series of words from within individual sentences;
for at least some of the small chunks, in a process separate from identifying the small chunks, automatically selecting text from sentences surrounding the small chunks as labels for semantic roles played by the small chunks in a transaction described by the individual documents;
standardizing the labels for semantic roles across the documents in the document set; and
annotating documents in the document set wherein the annotations include locations of the identified small chunks and standardized labels for the semantic roles played by the identified small chunks at those locations.
19. A computer system for analyzing and improving documents, the computer system comprising:
a storage medium for receiving and storing a document set that contains a plurality of documents; and
a processor system having access to the storage medium and executing an application program for analyzing and annotating documents, wherein the processor system executing the application program:
imports documents in the document set;
performs visual extraction of the imported documents, including creating signatures for document parts;
automatically identifies a hierarchical structure of chunks within individual documents in the document set (a) based on the visual extraction, content, and contexts in the individual document; and (b) based on patterns of visual extraction and content, across the documents in the document set, wherein the hierarchical structure includes small chunks comprising series of words from within individual sentences;
for at least some of the small chunks, in a process separate from identifying the small chunks, automatically selects text from sentences surrounding the small chunks as labels for semantic roles played by the small chunks in a transaction described by the individual documents;
standardizes the labels for semantic roles across the documents in the document set; and
annotates documents in the document set wherein the annotations include locations of the identified small chunks and standardized labels for the semantic roles played by the identified small chunks at those locations.
20. The computer-implemented method of claim 1 , wherein importing the documents includes importing word-processor documents and importing page-layout documents.