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


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(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: Token-Level Supervised Contrastive Learning for Punctuation Restoration

Abstract: Punctuation is critical in understanding natural language text. Currently, most automatic speech recognition (ASR) systems do not generate punctuation, which affects the performance of downstream tasks, such as intent detection and slot filling. This gives rise to the need for punctuation restoration. Recent work in punctuation restoration heavily utilizes pre-trained language models without considering data imbalance when predicting punctuation classes. In this work, we address this problem by proposing a token-level supervised contrastive learning method that aims at maximizing the distance of representation of different punctuation marks in the embedding space. The result shows that training with token-level supervised contrastive learning obtains up to 3.2% absolute F1 improvement on the test set.
Comments: 5 pages, 3 figures
Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG)
DOI: 10.21437/Interspeech.2021-661
Cite as: arXiv:2107.09099 [cs.CL]
  (or arXiv:2107.09099v2 [cs.CL] for this version)

Submission history

From: Qiushi Huang [view email]
[v1] Mon, 19 Jul 2021 18:24:33 GMT (1013kb,D)
[v2] Thu, 23 Sep 2021 16:18:39 GMT (1248kb,D)

Link back to: arXiv, form interface, contact.