Use of machine learning to characterize reference relationship applied over a citation graph
Techniques for document analysis using machine learning are provided. A selection of an index is received document, and a plurality of documents that refer to the index document is identified. For each respective document in the plurality of documents, a respective portion of the respective document is extracted, where the respective portion refers to the index document, and a respective vector representation is generated for the respective portion. A plurality of groupings is generated for the plurality of documents based on how each of the plurality of documents relate to the index document, by processing the vector representations using a trained classifier. Finally, at least an indication of the plurality of groupings is provided, along with the index document.
1. A method comprising:
receiving a selection of an index document;
identifying a plurality of documents that refer to the index document;
for each respective document in the plurality of documents:
extracting a respective portion of the respective document, wherein the respective portion refers to the index document; and
generating, by operation of one or more processors, a respective vector representation for the respective portion;
generating a plurality of groupings for the plurality of documents based on how each of the plurality of documents relate to the index document, comprising, for each respective document in the plurality of documents:
processing the respective vector representation using a trained classifier to assign the respective document to a respective category of a plurality of categories, wherein the plurality of categories includes (i) a category for documents that support the index document, and (ii) a category for documents that do not support the index document; and
providing at least an indication of the plurality of groupings, along with the index document.
2. The method of claim 1 , wherein the trained classifier is trained using a curated training set of documents, wherein each respective document in the curated training set is labeled based on how it relates to a respective reference document.
3. The method of claim 1 , wherein identifying the plurality of documents that refer to the index document comprises:
identifying documents that cite the index document; and
identifying documents that refer to the index document without explicitly citing the index document.
4. The method of claim 1 , wherein extracting the respective portion of each respective document comprises:
identifying a location in the respective document where the index document is referenced; and
extracting a predefined number of sentences before and after the identified location.
5. The method of claim 1 , wherein each of the plurality of groupings correspond to a respective category of a plurality of categories, wherein the plurality of categories further includes at least one of:
(i) a category for documents expanding on the index document;
(ii) a category for documents criticizing the index document;
(iii) a category for documents showing similar findings as the index document;
(iv) a category for documents that failed to show similar findings as the index document; or
(v) a category for documents that rely on the index document for support.
6. The method of claim 1 , the method further comprising:
providing a number of documents that are included in each of the plurality of groupings;
providing a link to each of the plurality of documents; and
providing the respective portion of each respective document in the plurality of documents.
7. The method of claim 1 , wherein the documents included in a first category of the plurality of groupings are sorted based in part on a respective importance measure of each respective document.
8. The method of claim 1 , wherein identifying the plurality of documents that refer to the index document comprises accessing a citation graph that includes the index document.
9. A computer-readable storage medium having computer-readable program code embodied therewith, the computer-readable program code executable by one or more computer processors to perform an operation comprising:
receiving a selection of an index document;
identifying a plurality of documents that refer to the index document;
for each respective document in the plurality of documents:
extracting a respective portion of the respective document, wherein the respective portion refers to the index document; and
generating a respective vector representation for the respective portion;
generating a plurality of groupings for the plurality of documents based on how each of the plurality of documents relate to the index document, comprising, for each respective document in the plurality of documents:
processing the respective vector representation using a trained classifier to assign the respective document a respective category of a plurality of categories, wherein the plurality of categories includes (i) a category for documents that support the index document, and (ii) a category for documents that do not support the index document; and
providing at least an indication of the plurality of groupings, along with the index document.
10. The computer-readable storage medium of claim 9 , wherein the trained classifier is trained using a curated training set of documents, wherein each respective document in the curated training set is labeled based on how it relates to a respective reference document.
11. The computer-readable storage medium of claim 9 , wherein identifying the plurality of documents that refer to the index document comprises:
identifying documents that cite the index document; and
identifying documents that refer to the index document without explicitly citing the index document.
12. The computer-readable storage medium of claim 9 , wherein extracting the respective portion of each respective document comprises:
identifying a location in the respective document where the index document is referenced; and
extracting a predefined number of sentences before and after the identified location.
13. The computer-readable storage medium of claim 9 , wherein each of the plurality of groupings correspond to a respective category of a plurality of categories, wherein the plurality of categories further includes at least one of:
(i) a category for documents expanding on the index document;
(ii) a category for documents criticizing the index document;
(iii) a category for documents showing similar findings as the index document;
(iv) a category for documents that failed to show similar findings as the index document; or
(v) a category for documents that rely on the index document for support.
14. The computer-readable storage medium of claim 9 , the operation further comprising:
providing a number of documents that are included in each of the plurality of groupings;
providing a link to each of the plurality of documents; and
providing the respective portion of each respective document in the plurality of documents.
15. A system comprising:
one or more computer processors; and
a memory containing a program which when executed by the one or more computer processors performs an operation, the operation comprising:
receiving a selection of an index document;
identifying a plurality of documents that refer to the index document;
for each respective document in the plurality of documents:
extracting a respective portion of the respective document, wherein the respective portion refers to the index document; and
generating a respective vector representation for the respective portion;
generating a plurality of groupings for the plurality of documents based on how each of the plurality of documents relate to the index document, comprising, for each respective document in the plurality of documents:
processing the respective vector representation using a trained classifier to assign the respective document a respective category of a plurality of categories, wherein the plurality of categories includes (i) a category for documents that support the index document, and (ii) a category for documents that do not support the index document; and
providing at least an indication of the plurality of groupings, along with the index document.
16. The system of claim 15 , wherein the trained classifier is trained using a curated training set of documents, wherein each respective document in the curated training set is labeled based on how it relates to a respective reference document.
17. The system of claim 15 , wherein identifying the plurality of documents that refer to the index document comprises:
identifying documents that cite the index document; and
identifying documents that refer to the index document without explicitly citing the index document.
18. The system of claim 15 , wherein extracting the respective portion of each respective document comprises:
identifying a location in the respective document where the index document is referenced; and
extracting a predefined number of sentences before and after the identified location.
19. The system of claim 15 , wherein each of the plurality of groupings correspond to a respective category of a plurality of categories, wherein the plurality of categories further includes at least one of:
(i) a category for documents expanding on the index document;
(ii) a category for documents criticizing the index document;
(iii) a category for documents showing similar findings as the index document;
(iv) a category for documents that failed to show similar findings as the index document; or
(v) a category for documents that rely on the index document for support.
20. The system of claim 15 , the operation further comprising:
providing a number of documents that are included in each of the plurality of groupings;
providing a link to each of the plurality of documents; and
providing the respective portion of each respective document in the plurality of documents.