Further Progress in Meeting Recognition: The ICSI-SRI Spring 2005 Speech-to-Text Evaluation System

Stolcke, Andreas; Anguera, Xavier; Boakye, Kofi; Çetin, Özgür; Grézl, František; Janin, Adam; Mandal, Arindam; Peskin, Barbara; Wooters, Chuck; Zheng, Jing

doi:10.1007/11677482_39

Cited by 45 publications

(39 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As expected, the background noise does not introduce any additional insertions over the clean condition, and most additional errors in the cross-talk condition are insertions and deletions (which tend to associate with each other). Such a dramatic increase in insertions and deletions for the cross-talk condition is in agreement with the results in [8] for real-world cross-talk, and provides a sanity check for the design.…”

Section: Wersupporting

confidence: 83%

“…Recognition experiments are conducted using the 2005 ICSI-SRI meeting system [8]. This system is adapted from SRI's conversational telephone speech system to the meeting domain using a variety of meeting data (including about 72 hours from the ICSI meeting corpus, excluding our test data).…”

Section: Recognition Systemmentioning

confidence: 99%

“…N -gram language models (LMs) with order as high as four were trained on standard text and meeting transcriptions as well as on Web texts. See [8] for full details. To avoid confounds with automatic speech segmentation, we use manual reference segmentations in our experiments, as our goal is to study the effects of overlaps on automatic speech recognition.…”

Section: Recognition Systemmentioning

confidence: 99%

See 2 more Smart Citations

Speaker Overlaps and ASR Errors in Meetings: Effects Before, During, and After the Overlap

Çetin

Shriberg

2006 IEEE International Conference on Acoustics Speed and Signal Processing Proceedings

View full text Add to dashboard Cite

We analyze automatic speech recognition (ASR) errors made by a state-of-the-art meeting recognizer, with respect to locations of overlapping speech. Our analysis focuses on recognition errors made both during an overlap and in the regions immediately preceding and following the location of overlapped speech. We devise an experimental paradigm to allow examination of the same foreground speech both with and without naturally occurring cross-talk. We then analyze ASR errors with respect to a number of factors, including the severity of the cross-talk and distance from the overlap region. In addition to reporting effects on ASR errors, we discover a number of interesting phenomena. First, we find that overlaps tend to occur at high-perplexity regions in the foreground talker's speech. Second, word sequences within overlaps have higher perplexity than those in nonoverlaps, if using trigrams or 4-grams, but the unigram perplexity within overlaps is considerably lower than that of nonoverlaps. An explanation for this behavior is proposed, based on the preponderance of multiple short dialog acts found in overlap regions. Third, we discover that the word error rate (WER) after overlaps is consistently lower than that before the overlap. This finding cannot be explained by the recognition process itself; rather, the foreground speaker appears to reduce perplexity shortly after being overlapped. Taken together, these observations suggest that the automatic modeling of meetings could benefit from a broader view of the relationship between speaker overlap and ASR in natural conversation.

show abstract

Section: Wersupporting

confidence: 83%

Section: Recognition Systemmentioning

confidence: 99%

Section: Recognition Systemmentioning

confidence: 99%

See 1 more Smart Citation

Speaker Overlaps and ASR Errors in Meetings: Effects Before, During, and After the Overlap

Çetin

Shriberg

2006 IEEE International Conference on Acoustics Speed and Signal Processing Proceedings

View full text Add to dashboard Cite

show abstract

“…The amount of speech resources for meetings is still relatively small and most systems make use of adaptation of models from other domains. In [13], a recognition system for conversational telephone speech (CTS) formed the starting point, others have reported that bootstrapping from Broadcast News (BN) systems works well.…”

Section: Meeting Speech Recognitionmentioning

confidence: 99%

Recognition and understanding of meetings the AMI and AMIDA projects

Renals

Hain

Bourlard

2007

2007 IEEE Workshop on Automatic Speech Recognition &Amp; Understanding (ASRU)

111

View full text Add to dashboard Cite

The AMI and AMIDA projects are concerned with the recognition and interpretation of multiparty meetings. Within these projects we have: developed an infrastructure for recording meetings using multiple microphones and cameras; released a 100 hour annotated corpus of meetings; developed techniques for the recognition and interpretation of meetings based primarily on speech recognition and computer vision; and developed an evaluation framework at both component and system levels. In this paper we present an overview of these projects, with an emphasis on speech recognition and content extraction.

show abstract

“…There has been extensive research at ICSI in the last few years in the area of speaker segmentation and diarization [1,2,3,4,5,6,7]. Speaker diarization is the task of identifying the number of participants in a meeting and create a list of speech time intervals for each such participant.…”

Section: Introductionmentioning

confidence: 99%

Speaker Diarization for Multi-microphone Meetings Using Only Between-Channel Differences

Pardo

Anguera

Wooters

2006

Machine Learning for Multimodal Interaction

Self Cite

View full text Add to dashboard Cite

Abstract. We present a method to extract speaker turn segmentation from multiple distant microphones (MDM) using only delay values found via a crosscorrelation between the available channels. The method is robust against the number of speakers (which is unknown to the system), the number of channels, and the acoustics of the room. The delays between channels are processed and clustered to obtain a segmentation hypothesis. We have obtained a 31.2% diarization error rate (DER) for the NIST´s RT05s MDM conference room evaluation set. For a MDM subset of NIST´s RT04s development set, we have obtained 36.93% DER and 35.73% DER * . Comparing those results with the ones presented by Ellis and Liu [8], who also used between-channels differences for the same data, we have obtained 43% relative improvement in the error rate.

show abstract

Further Progress in Meeting Recognition: The ICSI-SRI Spring 2005 Speech-to-Text Evaluation System

Cited by 45 publications

References 12 publications

Speaker Overlaps and ASR Errors in Meetings: Effects Before, During, and After the Overlap

Speaker Overlaps and ASR Errors in Meetings: Effects Before, During, and After the Overlap

Recognition and understanding of meetings the AMI and AMIDA projects

Speaker Diarization for Multi-microphone Meetings Using Only Between-Channel Differences

Contact Info

Product

Resources

About