Large Vocabulary Continuous Speech Recognition 1
HomeFull List of Titles
1: ICSLP'98 Proceedings
Keynote SpeechesText-To-Speech Synthesis 1Spoken Language Models and Dialog 1Prosody and Emotion 1Hidden Markov Model Techniques 1Speaker and Language Recognition 1Multimodal Spoken Language Processing 1Isolated Word RecognitionRobust Speech Processing in Adverse Environments 1Spoken Language Models and Dialog 2Articulatory Modelling 1Talking to Infants, Pets and LoversRobust Speech Processing in Adverse Environments 2Spoken Language Models and Dialog 3Speech Coding 1Articulatory Modelling 2Prosody and Emotion 2Neural Networks, Fuzzy and Evolutionary Methods 1Utterance Verification and Word Spotting 1 / Speaker Adaptation 1Text-To-Speech Synthesis 2Spoken Language Models and Dialog 4Human Speech Perception 1Robust Speech Processing in Adverse Environments 3Speech and Hearing Disorders 1Prosody and Emotion 3Spoken Language Understanding Systems 1Signal Processing and Speech Analysis 1Spoken Language Generation and Translation 1Spoken Language Models and Dialog 5Segmentation, Labelling and Speech Corpora 1Multimodal Spoken Language Processing 2Prosody and Emotion 4Neural Networks, Fuzzy and Evolutionary Methods 2Large Vocabulary Continuous Speech Recognition 1Speaker and Language Recognition 2Signal Processing and Speech Analysis 2Prosody and Emotion 5Robust Speech Processing in Adverse Environments 4Segmentation, Labelling and Speech Corpora 2Speech Technology Applications and Human-Machine Interface 1Large Vocabulary Continuous Speech Recognition 2Text-To-Speech Synthesis 3Language Acquisition 1Acoustic Phonetics 1Speaker Adaptation 2Speech Coding 2Hidden Markov Model Techniques 2Multilingual Perception and Recognition 1Large Vocabulary Continuous Speech Recognition 3Articulatory Modelling 3Language Acquisition 2Speaker and Language Recognition 3Text-To-Speech Synthesis 4Spoken Language Understanding Systems 4Human Speech Perception 2Large Vocabulary Continuous Speech Recognition 4Spoken Language Understanding Systems 2Signal Processing and Speech Analysis 3Human Speech Perception 3Speaker Adaptation 3Spoken Language Understanding Systems 3Multimodal Spoken Language Processing 3Acoustic Phonetics 2Large Vocabulary Continuous Speech Recognition 5Speech Coding 3Language Acquisition 3 / Multilingual Perception and Recognition 2Segmentation, Labelling and Speech Corpora 3Text-To-Speech Synthesis 5Spoken Language Generation and Translation 2Human Speech Perception 4Robust Speech Processing in Adverse Environments 5Text-To-Speech Synthesis 6Speech Technology Applications and Human-Machine Interface 2Prosody and Emotion 6Hidden Markov Model Techniques 3Speech and Hearing Disorders 2 / Speech Processing for the Speech and Hearing Impaired 1Human Speech ProductionSegmentation, Labelling and Speech Corpora 4Speaker and Language Recognition 4Speech Technology Applications and Human-Machine Interface 3Utterance Verification and Word Spotting 2Large Vocabulary Continuous Speech Recognition 6Neural Networks, Fuzzy and Evolutionary Methods 3Speech Processing for the Speech-Impaired and Hearing-Impaired 2Prosody and Emotion 72: SST Student Day
SST Student Day - Poster Session 1SST Student Day - Poster Session 2Author Index
A B C D E F G H I
J K L M N O P Q R
S T U V W X Y Z
Multimedia Files
Real-Time Recognition of Broadcast News
Authors:
Gary Cook, Cambridge University (U.K.)
Tony Robinson, Cambridge University (U.K.)
James Christie, Cambridge University (U.K.)
Page (NA) Paper number 65
Abstract:
Although the performance of state-of-the-art automatic speech recognition systems on the challenging task of broadcast news transcription has improved considerably in recent years, many of the systems operate in 130-300 times real-time. Many applications of automatic transcription of broadcast news, eg. closed-caption subtitles for television broadcasts, require real-time operation. This paper describes a connectionist-HMM system for broadcast news transcription, and the modifications to this system necessary for real-time operation. We show that real-time operation is possible with a relative increase in word error rate of about 12%.
SL980065.PDF (From Author)
SL980065.PDF (Rasterized)
TOP
Automatic Recognition of Korean Broadcast News Speech
Authors:
Ha-Jin Yu, LG Corporate Institute Of Technology (Korea)
Hoon Kim, LG Corporate Institute Of Technology (Korea)
Jae-Seung Choi, LG Corporate Institute Of Technology (Korea)
Joon-Mo Hong, LG Corporate Institute of Technology (Korea)
Kew-Suh Park, LG Corporate Institute of Technology (Korea)
Jong-Seok Lee, LG Corporate Institute of Technology (Korea)
Hee-Youn Lee, LG Corporate Institute of Technology (Korea)
Page (NA) Paper number 412
Abstract:
This paper describes preliminary results of automatic recognition of Korean broadcast-news speech. We have been working on flexible vocabulary isolated-word speech recognition, and the same HMM models are used for broadcast-news continuous speech recognition. The recognizer is trained by using phonetically balanced isolated words speech, rather than the broadcast news speech itself. In this research, we use several different lexica to investigate the recognition performance according to the length of the words. We also propose a long-distance bigram language model, which can be used at the first stage of the search, so that it can reduce the recognition errors caused by earlier pruning of correct hypothesis.
SL980412.PDF (From Author)
SL980412.PDF (Rasterized)
TOP
Telephone-Based Conversational Speech Recognition in the JUPITER Domain
Authors:
James R. Glass, MIT Lab for Computer Science (USA)
Timothy J. Hazen, MIT Lab for Computer Science (USA)
Page (NA) Paper number 593
Abstract:
This paper describes our experiences with developing a telephone-based speech recognizer as part of a conversational system in the weather information domain. This system has been used to collect spontaneous speech data which has proven to be extremely valuable for research in a number of different areas. After describing the corpus we have collected, we describe the development of the recognizer vocabulary, pronunciations, language and acoustic models for this system, and report on the current performance of the recognizer under several different conditions.
SL980593.PDF (From Author)
SL980593.PDF (Rasterized)
TOP
Japanese Large-Vocabulary Continuous Speech Recognition System Based on Microsoft Whisper
Authors:
Hsiao-Wuen Hon, Microsoft Research (USA)
Yun-Cheng Ju, Microsoft Research (USA)
Keiko Otani, Microsoft Research (USA)
Page (NA) Paper number 597
Abstract:
Input of Asian ideographic characters has traditionally been one of the biggest impediments for information processing in Asia. Speech is arguably the most effective and efficient input method for Asian non-spelling characters. This paper presents a Japanese large-vocabulary continuous speech recognition system based on Microsoft Whisper technology. We focus on the aspects of the system that are language specific and demonstrate the adaptability of the Whisper system to new languages. In this paper, we demonstrate that our pronunciation/part-of-speech distinguished morpheme based language models and Whisper based Japanese senonic acoustic models are able to yield state-of-the-art Japanese LVCSR recognition performance. The speaker-independent character and Kana error rates on the JNAS database are 10% and 5% respectively.
SL980597.PDF (From Author)
SL980597.PDF (Rasterized)
TOP
Partitioning And Transcription Of Broadcast News Data
Authors:
Jean-Luc Gauvain, LIMSI/CNRS (France)
Lori F. Lamel, LIMSI/CNRS (France)
Gilles Adda, LIMSI/CNRS (France)
Page (NA) Paper number 84
Abstract:
Radio and television broadcasts consist of a continuous stream of data comprised of segments of different linguistic and acoustic natures, which poses challenges for transcription. In this paper we report on our recent work in transcribing broadcast news data, including the problem of partitioning the data into homogeneous segments prior to word recognition. Gaussian mixture models are used to identify speech and non-speech segments. A maximum-likelihood segmentation/clustering process is then applied to the speech segments using GMMs and an agglomerative clustering algorithm. The clustered segments are then labeled according to bandwidth and gender. The recognizer is a continuous mixture density, tied-state cross-word context-dependent HMM system with a 65k trigram language model. Decoding is carried out in three passes, with a final pass incorporating cluster-based test-set MLLR adaptation. The overall word transcription error on the Nov'97 unpartitioned evaluation test data was 18.5%.