Automatic indexing of key sentences for lecture archives

Kawahara, Tatsuya; Shitaoka, Kazuya; Kitade, Tasuku; Nanjo, H.

doi:10.1109/asru.2003.1318418

Cited by 3 publications

(3 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Using a 5000 word vocabulary and trigram language model (perplexity 120) derived from a portion of lecture transcriptions and text book, we obtained a 33% word error rate on unseen lectures. This result is in line with other lecture word error rates of 30-40% that have been reported in the literature (Leeuwis et al, 2003;Kawahara et al, 2003).…”

Section: Preliminary Transcript Generationsupporting

confidence: 93%

“…The speech recognition processing that has been used to generate transcripts of spoken lectures has largely been based on large-vocabulary continuous speech recognition technology (Hurst et al, 2002;Leeuwis et al, 2003;Kawahara et al, 2003;Yokoyama et al, 2003). Language modeling research has focused on mixing topicdependent textual source material (e.g., conference papers) with unrelated or topic-independent spoken material (e.g., Switchboard data, or transcripts of other spoken material) (Kato et al, 2000).…”

Section: Preliminary Transcript Generationmentioning

confidence: 99%

“…Only 6 of the presenters were native English speakers however, and only 39 of the lectures have been transcribed. The Corpus of Spontaneous Japanese currently contains over 2,500 transcribed presentations (Kawahara et al, 2003). Both of these corpora focus on conference presentations, which are shorter and have a lower degree of spontaneity than a one hour or 90 minute classroom lecture.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Analysis and processing of lecture audio data

Glass

Hazen

Hetherington

et al. 2004

Proceedings of the Workshop on Interdisciplinary Approaches to Speech Indexing and Retrieval at HLT-NAACL 2004 - SpeechIR '04

View full text Add to dashboard Cite

In this paper we report on our recent efforts to collect a corpus of spoken lecture material that will enable research directed towards fast, accurate, and easy access to lecture content. Thus far, we have collected a corpus of 270 hours of speech from a variety of undergraduate courses and seminars. We report on an initial analysis of the spontaneous speech phenomena present in these data and the vocabulary usage patterns across three courses. Finally, we examine language model perplexities trained from written and spoken materials, and describe an initial recognition experiment on one course.

show abstract

Section: Preliminary Transcript Generationsupporting

confidence: 93%

Section: Preliminary Transcript Generationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Analysis and processing of lecture audio data

Glass

Hazen

Hetherington

et al. 2004

Proceedings of the Workshop on Interdisciplinary Approaches to Speech Indexing and Retrieval at HLT-NAACL 2004 - SpeechIR '04

View full text Add to dashboard Cite

show abstract

Automatic extraction of key sentences from oral presentations using statistical measure based on discourse markers

Kitade¹,

Kawahara²,

Nanjo³

2004

Interspeech 2004

View full text Add to dashboard Cite

Real-Time Presentation Tracking Using Semantic Keyword Spotting

et al. 2016

View full text Add to dashboard Cite

Given presentation slides with detailed written speaking notes, automatic tracking of oral presentations can help speakers ensure they cover their planned content, and can reduce their anxiety during the speech. Tracking is a more complex problem than speech-to-text alignment, since presenters rarely follow their exact presentation notes, and it must be performed in realtime. In this paper, we propose a novel system that can track the current degree of coverage of each slide's contents. To do this, the presentation notes for each slide are segmented into sentences, and the words are filtered into keyword candidates. These candidates are then scored based on word specificity and semantic similarity measures to find the most useful keywords for the tracking task. Real-time automatic speech recognition results are matched against the keywords and their synonyms. Sentences are scored based on detected keywords, and the ones with scores higher than a threshold are tagged as covered. We manually and automatically annotated 150 slide presentation recordings to evaluate the system. A simple tracking method, matching speech recognition results against the notes, was used as the baseline. The results show that our approach led to higher accuracy measures compared to the baseline method.

show abstract

Automatic indexing of key sentences for lecture archives

Cited by 3 publications

References 6 publications

Analysis and processing of lecture audio data

Analysis and processing of lecture audio data

Automatic extraction of key sentences from oral presentations using statistical measure based on discourse markers

Real-Time Presentation Tracking Using Semantic Keyword Spotting

Contact Info

Product

Resources

About