We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CL

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Few-Shot Keyword Spotting in Any Language

Abstract: We introduce a few-shot transfer learning method for keyword spotting in any language. Leveraging open speech corpora in nine languages, we automate the extraction of a large multilingual keyword bank and use it to train an embedding model. With just five training examples, we fine-tune the embedding model for keyword spotting and achieve an average F1 score of 0.75 on keyword classification for 180 new keywords unseen by the embedding model in these nine languages. This embedding model also generalizes to new languages. We achieve an average F1 score of 0.65 on 5-shot models for 260 keywords sampled across 13 new languages unseen by the embedding model. We investigate streaming accuracy for our 5-shot models in two contexts: keyword spotting and keyword search. Across 440 keywords in 22 languages, we achieve an average streaming keyword spotting accuracy of 85.2% with a false acceptance rate of 1.2%, and observe promising initial results on keyword search.
Comments: Submitted to INTERSPEECH 2021
Subjects: Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
Cite as: arXiv:2104.01454 [cs.CL]
  (or arXiv:2104.01454v3 [cs.CL] for this version)

Submission history

From: Mark Mazumder [view email]
[v1] Sat, 3 Apr 2021 17:27:37 GMT (3274kb,D)
[v2] Tue, 6 Apr 2021 15:48:01 GMT (3274kb,D)
[v3] Thu, 22 Apr 2021 18:58:44 GMT (3274kb,D)
[v4] Thu, 9 Sep 2021 20:36:28 GMT (4459kb,D)

Link back to: arXiv, form interface, contact.