Explainable unsupervised vector representation of multi-section documents
Embodiments of the present disclosure provide methods, apparatus, systems, computing devices, computing entities, and/or the like for generating an inferred document representation for a multi-section document using a machine learning model. In accordance with one embodiment, a method is provided that includes: identifying a document corpus comprising the multi-section document and other multi-section documents; for each section of the document that is associated with a section type identifier: identifying a section batch that comprises common-type sections across the document corpus; and processing the section batch using the machine learning model to generate per-type section clusters for the section type identifier that comprise an inferred per-type section cluster for the current section; generating the inferred document representation based at least in part on each inferred per-type section cluster for a section of the document; and performing a prediction-based action based at least in part on the representation.
1. A computer-implemented method for automated generation of an inferred document representation for a multi-section document to convey detailed content granularity of semantic characteristics within the multi-section document, the computer-implemented method comprising:
identifying, by one or more processors, a document corpus comprising the multi-section document and a plurality of other multi-section documents, wherein the multi-section document comprises a plurality of sections that are respectively associated with a plurality of document-wide section type identifiers;
identifying, by the one or more processors, a section batch from a current section of the plurality of sections that is associated with a current section type identifier from the plurality of document-wide section type identifiers, wherein:
(i) the section batch comprises a plurality of common-type sections that is associated with the current section type identifier,
(ii) the plurality of common-type sections comprises the current section and one or more other sections that are associated with the current section type identifier across the plurality of other multi-section documents, and
(iii) the current section type identifier corresponds to one or more semantic characteristics of the plurality of common-type sections;
inputting, by the one or more processors, the section batch to an unsupervised section clustering machine learning model to generate a plurality of per-type section clusters for the current section type identifier, wherein:
(i) each per-type section cluster of the plurality of per-type section clusters comprises a related section subset of the plurality of common-type sections, and
(ii) the plurality of per-type section clusters comprises an inferred per-type section cluster for the current section;
generating, by the one or more processors, the inferred document representation for the multi-section document based at least in part on a plurality of inferred per-type section clusters, including the inferred per-type section cluster, respectively generated for the plurality of sections; and
initiating, by the one or more processors, performance of one or more prediction-based actions based at least in part on the inferred document representation.
2. The computer-implemented method of claim 1 , further comprising:
identifying, by the one or more processors, a section schema associated with the document corpus, wherein: (i) the section schema describes a group of corpus-wide section type identifiers for the document corpus, and (ii) the group of corpus-wide section type identifiers comprise the plurality of document-wide section type identifiers.
3. The computer-implemented method of claim 2 , wherein:
the inferred document representation describes a plurality of per-type section cluster identifiers;
a per-type section cluster identifier of the plurality of per-type section cluster identifiers is associated with a corpus-wide section type identifier of the group of corpus-wide section type identifiers;
the per-type section cluster identifier for the corpus-wide section type identifier of the group of corpus-wide section type identifiers that is among the plurality of document-wide section type identifiers describes the inferred per-type section cluster for the current section of the plurality of sections that is associated with the corpus-wide section type identifier; and
the per-type section cluster identifier for the corpus-wide section type identifier of the group of corpus-wide section type identifiers that is not among the plurality of document-wide section type identifiers describes a default numerical value.
4. The computer-implemented method of claim 1 , further comprising:
processing, by the one or more processors, a per-type section cluster of the plurality of per-type section clusters using a document summarization machine learning model to generate a per-type section cluster summary for the per-type section cluster.
5. The computer-implemented method of claim 4 , wherein generating the per-type section cluster summary for the per-type section cluster of the plurality of per-type section clusters comprises:
processing the related section subset for the per-type section cluster using the document summarization machine learning model to generate the per-type section cluster summary.
6. The computer-implemented method of claim 4 , wherein performing the one or more prediction-based actions comprises:
causing presentation of a prediction output user interface, wherein: (i) the prediction output user interface describes a multi-section document summary for the multi-section document, and (ii) the multi-section document summary describes each the per-section type cluster summary for the per-type section cluster of the plurality of per-type section clusters.
7. The computer-implemented method of claim 1 , wherein performing the one or more prediction-based actions comprises:
causing presentation of a prediction output user interface, wherein the prediction output user interface describes the inferred document representation.
8. The computer-implemented method of claim 7 , wherein the prediction output user interface describes a multi-section document summary for the multi-section document.
9. A system for automated generation of an inferred document representation for a multi-section document to convey detailed content granularity of semantic characteristics within the multi-section document, the system comprising memory and one or more processors communicatively coupled to the memory, the one or more processors configured to:
identify a document corpus comprising the multi-section document and a plurality of other multi-section documents, wherein the multi-section document comprises a plurality of sections that are respectively associated with a plurality of document-wide section type identifiers;
identify a section batch from a current section of the plurality of sections that is associated with a current section type identifier from the plurality of document-wide section type identifiers, wherein:
(i) the section batch comprises a plurality of common-type sections that are associated with the current section type identifier,
(ii) the plurality of common-type sections comprises the current section and one or more other sections that are associated with the current section type identifier across the plurality of other multi-section documents, and
(iii) the current section type identifier corresponds to one or more semantic characteristics of the plurality of common-type sections;
input the section batch to an unsupervised section clustering machine learning model to generate a plurality of per-type section clusters for the current section type identifier, wherein:
(i) each per-type section cluster of the plurality of per-type section clusters comprises a related section subset of the plurality of common-type sections, and
(ii) the plurality of per-type section clusters comprises an inferred per-type section cluster for the current section;
generate the inferred document representation for the multi-section document based at least in part on a plurality of inferred per-type section clusters, including the inferred per-type section cluster, respectively generated for the plurality of sections; and
initiate performance of one or more prediction-based actions based at least in part on the inferred document representation.
10. The system of claim 9 , wherein the memory and the one or more processors are configured to:
identify a section schema associated with the document corpus, wherein: (i) the section schema describes a group of corpus-wide section type identifiers for the document corpus, and (ii) the group of corpus-wide section type identifiers comprise the plurality of document-wide section type identifiers.
11. The system of claim 10 , wherein:
the inferred document representation describes a plurality of per-type section cluster identifiers;
a per-type section cluster identifier of the plurality of per-type section cluster identifiers is associated with a corpus-wide section type identifier of the group of corpus-wide section type identifiers;
the per-type section cluster identifier for the corpus-wide section type identifier of the group of corpus-wide section type identifiers that is among the plurality of document-wide section type identifiers describes the inferred per-type section cluster for the current section of the plurality of sections that is associated with the corpus-wide section type identifier; and
the per-type section cluster identifier for the corpus-wide section type identifier of the group of corpus-wide section type identifiers that is not among the plurality of document-wide section type identifiers describes a default numerical value.
12. The system of claim 9 , wherein the one or more processors are further configured to:
process a per-type section cluster of the plurality of per-type section clusters using a document summarization machine learning model to generate a per-type section cluster summary for the per-type section cluster.
13. The system of claim 12 , wherein the one or more processors are further configured to:
process the related section subset for the per-type section cluster using the document summarization machine learning model to generate the per-type section cluster summary.
14. The system of claim 12 , wherein the one or more processors are further configured to:
cause presentation of a prediction output user interface, wherein: (i) the prediction output user interface describes a multi-section document summary for the multi-section document, and (ii) the multi-section document summary describes each per-section type cluster summary for the per-type section cluster of the plurality of per-type section clusters.
15. One or more non-transitory computer-readable storage media for automated generation of an inferred document representation for a multi-section document to convey detailed content granularity of semantic characteristics within the multi-section document, the one or more non-transitory computer-readable storage media including instructions that, when executed by one or more processors, cause the one or more processors to:
identify a document corpus comprising the multi-section document and a plurality of other multi-section documents, wherein the multi-section document comprises a plurality of sections that are respectively associated with a plurality of document-wide section type identifiers;
identify a section batch from a current section of the plurality of sections that is associated with a current section type identifier from the plurality of document-wide section type identifiers, wherein:
(i) the section batch comprises a plurality of common-type sections that is associated with the current section type identifier,
(ii) the plurality of common-type sections comprises the current section and one or more other sections that are associated with the current section type identifier across the plurality of other multi-section documents, and
(iii) the current section type identifier corresponds to one or more semantic characteristics of the plurality of common-type sections;
input the section batch to an unsupervised section clustering machine learning model to generate a plurality of per-type section clusters for the current section type identifier, wherein:
(i) each per-type section cluster of the plurality of per-type section clusters comprises a related section subset of the plurality of common-type sections, and
(ii) the plurality of per-type section clusters comprises an inferred per-type section cluster for the current section;
generate the inferred document representation for the multi-section document based at least in part on a plurality of inferred per-type section clusters, including the inferred per-type section cluster, respectively generated for the plurality of sections; and
initiate performance of one or more prediction-based actions based at least in part on the inferred document representation.
16. The one or more non-transitory computer-readable storage media of claim 15 , further including instructions that, when executed by the one or more processors, cause the one or more processors to:
identify a section schema associated with the document corpus, wherein: (i) the section schema describes a group of corpus-wide section type identifiers for the document corpus, and (ii) the group of corpus-wide section type identifiers comprise the plurality of document-wide section type identifiers.
17. The one or more non-transitory computer-readable storage media of claim 16 , wherein:
the inferred document representation describes a plurality of per-type section cluster identifiers;
a per-type section cluster identifier of the plurality of per-type section cluster identifiers is associated with a corpus-wide section type identifier of the group of corpus-wide section type identifiers;
the per-type section cluster identifier for the corpus-wide section type identifier of the group of corpus-wide section type identifiers that is among the plurality of document-wide section type identifiers describes the inferred per-type section cluster for the current section of the plurality of sections that is associated with the corpus-wide section type identifier; and
the per-type section cluster identifier for the corpus-wide section type identifier of the group of corpus-wide section type identifiers that is not among the plurality of document-wide section type identifiers describes a default numerical value.
18. The one or more non-transitory computer-readable storage media of claim 15 , further including instructions that, when executed by the one or more processors, cause the one or more processors to:
process a per-type section cluster of the plurality of per-type section clusters using a document summarization machine learning model to generate a per-type section cluster summary for the per-type section cluster.
19. The one or more non-transitory computer-readable storage media of claim 18 , further including instructions that, when executed by the one or more processors, cause the one or more processors to:
process the related section subset for the per-type section cluster using the document summarization machine learning model to generate the per-type section cluster summary.
20. The one or more non-transitory computer-readable storage media of claim 18 , further including instructions that, when executed by the one or more processors, cause the one or more processors to:
cause presentation of a prediction output user interface, wherein: (i) the prediction output user interface describes a multi-section document summary for the multi-section document, and (ii) the multi-section document summary describes each per-section type cluster summary for the per-type section cluster of the plurality of per-type section clusters.