US 12,462,171 B2
Hierarchical context tagging for utterance rewriting
Linfeng Song, Palo Alto, CA (US)
Assigned to TENCENT AMERICA LLC, Palo Alto, CA (US)
Filed by TENCENT AMERICA LLC, Palo Alto, CA (US)
Filed on Nov. 22, 2021, as Appl. No. 17/456,051.
Prior Publication US 2023/0162055 A1, May 25, 2023
Int. Cl. G10L 15/22 (2006.01); G06N 5/04 (2023.01); G10L 15/06 (2013.01); G10L 15/16 (2006.01); G10L 15/193 (2013.01)
CPC G06N 5/04 (2013.01) [G10L 15/063 (2013.01); G10L 15/16 (2013.01); G10L 15/193 (2013.01); G10L 15/22 (2013.01); G10L 2015/223 (2013.01); G10L 2015/228 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method performed by at least one processor of a multi-span tagger (MST) of hierarchical context tagging for utterance rewriting, the method comprising:
obtaining source tokens and context tokens;
encoding, using a Bidirectional Encoder Representations from Transforms (BERT) model, the source tokens and the context tokens to generate first source contextualized embeddings and first context contextualized embeddings;
tagging, using the first source contextualized embeddings, the source tokens with tags indicating a keep or delete action for each source token of the source tokens;
selecting, using the first context contextualized embeddings, a rule containing a sequence of one or more slots to insert before the each source token;
generating spans from the context tokens, each span corresponding to one of the one or more slots of the selected rule; and
performing, using the spans, utterance rewriting on a multi-turn dialogue to recover one or more coreferences on a latest turn of the multi-turn dialogue.