Fei Liu
Emory University, GA·Computer Science·Early-career
Recordings of meetings, telehealth visits, podcasts and livestreams are now routinely captioned, producing huge collections of written transcripts that few people have time to read. Automatic summarization, where software condenses a long text into a short one, could help.
- Fields
Computer Science
- Topics
natural language processing
automatic summarization
transcript summarization
- Grants
- 1 known grant, through May 2027
What they work on
- Uncover the hidden structure of informal, verbose transcripts of spontaneous speech, using deep neural networks and linguistic structure prediction to infer how topics and segments of a conversation relate to one another.
- Produce comprehensive summaries built on this induced document structure, improving the efficiency of content selection so users can navigate long audio and video transcripts and locate important topics.
- Establish an evaluation protocol for transcript summaries that combines intrinsic measures, rating the summary text itself, with extrinsic measures of how useful the summary is for a downstream task.
More about the research
But spoken language is hard for these tools: transcripts of spontaneous speech are informal, wordy, and lack the headings and paragraphs that organize written documents.
Why it matters
Organizations holding large archives of spoken content often cannot surface the important parts without labor-intensive manual note-taking. Telehealth and telemedicine providers, financial services, video-conferencing platforms, and podcast and livestream services could use more robust transcript summarization to share meeting minutes, take notes on recordings, and pinpoint topics of importance rather than being overwhelmed by available information.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
David Harwath
University of Texas at Austin, TX·Computer Science·Early-career
Develop attention-based neural network segmentation layers that discover units in raw speech audio, applied hierarchically to recover structure at multiple levels such as phones, words and phrases.
Rui Zhang
Pennsylvania State University, PA·Computer Science·Early-career
Develop controllable summarization models that satisfy fine-grained, compositional user preferences for both the content and the style of a summary, along with rigorous task definitions and evaluation metrics.
Frederik Hartmann
University of North Texas, TX·Social Science
Build artificial neural networks that learn to reconstruct proto-forms, the ancestor words of a language family, from their descendant words, using large datasets spanning many language families.
The school
Emory University: 70 labs on PI Pages, $1.2B a year on research, 286 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.