US 11,983,489 B1
Extractive summary generation by abstractive trained model
Natalie Bar Eliyahu, Petah Tikva (IL); Ido Farhi, Tel Aviv (IL); Adi Shalev, Herzliya (IL); and Oren Dar, Tel Aviv (IL)
Assigned to Intuit Inc., Mountain View, CA (US)
Filed by Intuit Inc., Mountain View, CA (US)
Filed on Jun. 23, 2023, as Appl. No. 18/213,808.
Int. Cl. G06F 40/166 (2020.01); G06F 40/279 (2020.01)
CPC G06F 40/166 (2020.01) [G06F 40/279 (2020.01)] 24 Claims
OG exemplary drawing
 
1. A method for generating a text summary, the method comprising:
processing an input text by a model that comprises a set of attention heads, wherein the model is trained to generate abstractive summaries of text documents;
identifying a subset of the attention heads;
for each attention head in the subset, identifying a portion from the input text that is used to generate the abstractive summary;
for each sentence of the input text, calculating a fractional size of an intersection of the portion and the respective sentence relative to the respective sentence;
determining a subset of the sentences, wherein the respective fractional size of each sentence in the subset meets a first threshold; and
generating an extractive summary of the input text from the subset of sentences.