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: Using BERT Embeddings to Model Word Importance in Conversational Transcripts for Deaf and Hard of Hearing Users

Abstract: Deaf and hard of hearing individuals regularly rely on captioning while watching live TV. Live TV captioning is evaluated by regulatory agencies using various caption evaluation metrics. However, caption evaluation metrics are often not informed by preferences of DHH users or how meaningful the captions are. There is a need to construct caption evaluation metrics that take the relative importance of words in a transcript into account. We conducted correlation analysis between two types of word embeddings and human-annotated labeled word-importance scores in existing corpus. We found that normalized contextualized word embeddings generated using BERT correlated better with manually annotated importance scores than word2vec-based word embeddings. We make available a pairing of word embeddings and their human-annotated importance scores. We also provide proof-of-concept utility by training word importance models, achieving an F1-score of 0.57 in the 6-class word importance classification task.
Comments: 5 pages, 3 tables, 1 figure
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
DOI: 10.18653/v1/2022.ltedi-1.5
Cite as: arXiv:2206.12368 [cs.CL]
  (or arXiv:2206.12368v1 [cs.CL] for this version)

Submission history

From: Akhter Al Amin [view email]
[v1] Fri, 24 Jun 2022 16:35:57 GMT (129kb,D)

Link back to: arXiv, form interface, contact.