Non-transitory computer-readable storage medium and system for generating an abstractive text summary of a document
There is provided a non-transitory storage medium and a system for generating an abstractive summary of a document using an abstractive machine learning algorithm (MLA). A document including a plurality of text sequences is received. An extractive summary of the document is generated, the extractive summary including a set of summary text sequences which is a subset of the plurality of text sequences. The abstractive MLA generates, based on the set of summary text sequences and at least a portion of the plurality of text sequences, an abstractive summary of the document including a set of abstractive text sequences, at least one abstractive text sequence not being included in the plurality of text sequences.
1. A non-transitory computer-readable storage medium storing instructions thereon, the instructions, upon being executed by at least one processor, cause the at least one processor to:
receive a document, the document comprising a plurality of text sequences;
generate, using an extractive machine learning algorithm (MLA) having been trained to generate extractive text summaries, an extractive summary of the document, the extractive summary comprising a set of extractive summary text sequences, the set of extractive summary text sequences being a subset of the plurality of text sequences; and
generate, using an abstractive MLA having been trained to generate abstractive text summaries based on at least the extractive text summaries, based on the set of extractive summary text sequences and at least a portion of the plurality of text sequences, an abstractive summary of the document comprising a set of abstractive summary text sequences, at least one abstractive summary text sequence not being included in the plurality of text sequences.
2. The non-transitory computer-readable storage medium of claim 1 , wherein the extractive MLA comprises one of a pointer network and a sentence classifier.
3. The non-transitory computer-readable storage medium of claim 2 , wherein the abstractive MLA comprises a transformer language model (TLM).
4. The non-transitory computer-readable storage medium of claim 3 , wherein the at least portion of the plurality of text sequences comprises an introduction section of the document.
5. The non-transitory computer-readable storage medium of claim 4 , wherein the abstractive MLA uses the extractive summary text sequences and at least the portion of the plurality of text sequences as conditioning to generate the abstractive summary.
6. The non-transitory computer-readable storage medium of claim 1 , wherein the document comprises above 1000 words.
7. The non-transitory computer-readable storage medium of claim 1 , wherein the document comprises one of: a news article, a web page, a scientific article, and a patent publication.
8. A system for generating an abstractive summary of a document, the system comprising:
a non-transitory storage medium storing computer-readable instructions; and
at least one processor operatively connected to the non-transitory storage medium, the at least one processor, upon executing the computer-readable instructions, being configured to cause:
receiving the document, the document comprising a plurality of text sequences;
generating an extractive summary of the document, the extractive summary comprising a set of summary text sequences, the set of summary text sequences being a subset of the plurality of text sequences; and
generating, by the abstractive MLA, based on the set of summary text sequences and at least a portion of the plurality of text sequences, an abstractive summary of the document comprising a set of abstractive text sequences, at least one abstractive text sequence not being included in the plurality of text sequences.
9. The system of claim 8 , wherein the extractive MLA comprises one of a pointer network and a sentence classifier.
10. The system of claim 9 , wherein the abstractive MLA comprises a transformer language model (TLM).
11. The system of claim 10 , wherein the at least portion of the plurality of text sequences comprises an introduction section of the document.
12. The system of claim 8 , wherein the abstractive MLA uses the extractive summary text sequences and at least the portion of the plurality of text sequences as conditioning to generate the abstractive summary.
13. The system of claim 8 , wherein the document comprises above 1000 words.
14. The system of claim 8 , wherein the document comprises one of: a news article, a web page, a scientific article and a patent publication.