Advances in meeting recognition

Waibel, Alex; Yu, Hua; Westphal, Martin; Soltau, Hagen; Schultz, Tanja; Schaaf, Thomas; Pan, Yingtian; Metze, Florian; Bett, M.

doi:10.3115/1072133.1072136

Cited by 17 publications

(10 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [9] a bigram language model is therefore used for computing the lookahead scores for a one-pass decoder using tree copies. However, the results in table 4 show, that the full incorporation of the language model for the lookahead tree is even useful for very hard tasks with weak language models as in the meeting scenario. In case of matched conditions, we found a speed-up of up to 26% by using the full lookahead.…”

Section: Resultsmentioning

confidence: 98%

“…The second task is read speech from the Broadcast News (BN) corpus; it consists of clean, read speech from a very large domain. The third task is a subset of the Meeting data set [4], which was recorded at informal group meetings, containing very colloquial speech recorded through lapel microphones. Decoding runs on the meeting data can be seen as a 'stress' test, since the acoustic and language models, trained on BN and ESST, don't match the test conditions, so that wide beam thresholds are needed to avoid search errors.…”

Section: Methodsmentioning

confidence: 99%

“…The results show that the proposed concept of polymorphic linguistic context assignment applied to a re-entrant pronunciation prefix tree is particularly effective for decoding with very large vocabularies. Smaller systems improve, too, albeit not as much as larger systems, which are receiving a lot of attention in the context of unrestricted tasks such as meeting recognition [4].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Efficient language model lookahead through polymorphic linguistic context assignment

Soltau¹,

Metze²,

Fügen³

et al. 2002

IEEE International Conference on Acoustics Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

In this study, we examine how fast decoding of conversational speech with large vocabularies profits from efficient use of linguistic information, i.e. language models and grammars. Based on a re-entrant single pronunciation prefix tree, we use the concept of linguistic context polymorphism to achieve an early incorporation of language model information. This approach allows us to use all available language model information in a one-pass decoder, using the same engine to decode with statistical n-gram language models as well as context free grammars or re-scoring of lattices in an efficient way. We compare this approach to our previous decoder, which needed three passes to incorporate all available information. The results on a very large vocabulary task show that the search can be speeded up by almost a factor of three, without introducing additional search errors. On all examined tasks, we observed significant improvements by using an exact language model lookahead over usual bigram lookahead strategies, even for very hard tasks with unmatched conditions, without introducing extra memory overhead.

show abstract

Section: Resultsmentioning

confidence: 98%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Efficient language model lookahead through polymorphic linguistic context assignment

Soltau¹,

Metze²,

Fügen³

et al. 2002

IEEE International Conference on Acoustics Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…With the doctor-patient data the drop in error rate was less severe which can be explained by the similar speaking style and recording conditions for C-STAR and doctor-patient data. Details about the applied recognition engine can be found in [10] for ESST and [11] for the BN system.…”

Section: Experiments 3: Porting the Speech Recognizer To New Domainsmentioning

confidence: 99%

Domain portability in speech-to-speech translation

Lavie

Levin

Schultz

et al. 2001

Proceedings of the First International Conference on Human Language Technology Research - HLT '01

Self Cite

View full text Add to dashboard Cite

“…An effective system will undoubtedly involve more than one search strategy. For retrieval of meetings, a variety of analysis techniques exist that extract keywords, topics, location, participant identity, etc [14]. There are also techniques that utilize turn-taking or other sequence information for meeting categorization.…”

Section: Related Workmentioning

confidence: 99%

Meeting Video Retrieval Using Dynamic HMM Model Similarity

Lee¹,

Hull²,

Erol³

2005 IEEE International Conference on Multimedia and Expo

View full text Add to dashboard Cite

Overcoming the semantic-feature gap and adapting to context are two main challenges in content-based retrieval. The problem is even more complicated for unstructured videos such as automated recordings of meetings. To address this problem, we propose a model-based approach to meeting retrieval with user controlled weighting for dynamic similarity comparison. Each video is represented by an HMM, and the similarity between videos is determined by comparing the corresponding models. Users can control the relative importance of temporal and static features by adjusting a weighting parameter in a way similar to content-based image retrieval. Experimental results demonstrate the feasibility and versatility of this approach.

show abstract

Advances in meeting recognition

Cited by 17 publications

References 7 publications

Efficient language model lookahead through polymorphic linguistic context assignment

Efficient language model lookahead through polymorphic linguistic context assignment

Domain portability in speech-to-speech translation

Meeting Video Retrieval Using Dynamic HMM Model Similarity

Contact Info

Product

Resources

About