Filters








26 Hits in 1.5 sec

The THISL spoken document retrieval project

S. Renals
Proceedings IEEE International Conference on Multimedia Computing and Systems  
In this paper we outline our spoken document retrieval system based on the ABBOT speech recognizer and a text retrieval system based on Okapi term-weighting .  ...  The system has been evaluated as part of the TREC-6 and TREC-7 spoken document retrieval evaluations and we report on the results of the TREC-7 evaluation based on a document collection of 100 hours of  ...  ACKNOWLEDGMENTS This work was supported by the ESPRIT Long Term Research Projects THISL (23495) and SPRACH (20077).  ... 
doi:10.1109/mmcs.1999.778655 dblp:conf/icmcs/Renals99 fatcat:w2kj4jma2vdkfgx5hr3xlj6htm

Robust spoken document retrieval methods for misrecognition and out-of-vocabulary keywords

Hiromitsu Nishizaki, Seiichi Nakagawa
2004 Systems and Computers in Japan  
This paper describes a Japanese spoken document retrieval (SDR) system that is robust for Out-of-Vocabulary (OOV) words.  ...  Evaluation results show that the proposed technique is quite effective in robustly retrieving spoken documents.  ...  Experiments in spoken document retrieval using Specification of models for LVCSR systems. The system structure for OOV keywords.  ... 
doi:10.1002/scj.10697 fatcat:liynm3th55hkpmexm7ovomr7qi

A speech interface for open-domain question-answering

Edward Schofield, Zhiping Zheng
2003 Proceedings of the 41st Annual Meeting on Association for Computational Linguistics - ACL '03  
of errors, and an open-domain question-answering system, AnswerBus, which is freely available on the Web.  ...  We describe a small evaluation of the effect of recognition errors on the precision of the answers returned and make some concrete recommendations for modifying a question-answering system for improving  ...  Acknowledgements The authors would like to thank Stefan Rüger for his suggestions and moral support. Ed Schofield's research is supported by a Marie Curie Fellowship of the European Commission.  ... 
doi:10.3115/1075178.1075210 dblp:conf/acl/SchofieldZ03 fatcat:hfbcjhjroraf7djzsurkctxc54

Statistical Language Modelling [chapter]

Yoshihiko Gotoh, Steve Renals
2003 Lecture Notes in Computer Science  
In a speech recognition system the role of the language model is to assign probabilities to word sequences.  ...  occurrence rate of a word is not uniform, but varies between documents.  ...  The Text Retrieval Conference (TREC) has been a forum for the evaluation of text retrieval systems for a variety of tasks including routing, filtering and spoken document retrieval.  ... 
doi:10.1007/978-3-540-45115-0_4 fatcat:tyxdr6nv7jbhxkxtndi7f7qkra

Indexing and retrieval of broadcast news

Steve Renals, Dave Abberley, David Kirby, Tony Robinson
2000 Speech Communication  
This paper describes a spoken document retrieval (SDR) system for British and North American Broadcast News.  ...  The system is based on a connectionist large vocabulary speech recognizer and a probabilistic information retrieval system.  ...  Acknowledgments This work was supported by ESPRIT Long Term Research Project THISL (EP23495). Thanks to Gary Cook for assistance with North American English broadcast speech  ... 
doi:10.1016/s0167-6393(00)00020-0 fatcat:wdj3zvhtkvb2zbbzcryqoqh2vy

Content-based access to spoken audio

K. Koumpis, S. Renals
2005 IEEE Signal Processing Magazine  
One way to facilitate retrieval is by classifying content into categories. The last and perhaps the least explored phase deals with the delivery of the retrieved content to users.  ...  We describe how the analysis, retrieval and delivery phases contribute making spoken audio content more accessible, and we outline a number of outstanding research issues.  ...  Koumpis and an MSc and a PhD from the University of Edinburgh. His research is in the areas of speech recognition, information access from spoken audio and models for multimodal data.  ... 
doi:10.1109/msp.2005.1511824 fatcat:a7p7ay3lmfen5brbmvqwoi6ete

Jan 2000 European Trip report: THISL and RESPITE

Daniel P. W. Ellis
2017
A report from meetings of the THISL (Thematic Indexing of Spoken Language) and RESPITE (Recognizing Speech by Partial Info. Techs.) European speech retrieval and recognition projects.  ...  for Aurora -other research, issues 1 2 THISL (Thematic Indexing of Spoken Language) • Spoken document retrieval of BBC Broadcast News -automatic off-air recording of 3-6 hrs daily news -ASR  ...  Outline -(GMM system does not know they are phones) Aurora "Distributed SR" evaluation • 7 telecoms company submissions:-Tandem systems from OGI-ICSI-Qualcomm• Best features for transmission?  ... 
doi:10.7916/d84x5h0p fatcat:lxriwfhftjdvllzxsakc74dj7e

Retrieval of broadcast news documents with the THISL system

D. Abberley, S. Renals, G. Cook
Proceedings of the 1998 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '98 (Cat. No.98CH36181)  
This paper describes the THISL system that participated in the TREC-7 evaluation, Spoken Document Retrieval (SDR) Track, and presents the results obtained, together with some analysis.  ...  The THISL system is based on the ABBOT speech recognition system and the thislIR text retrieval system.  ...  The THISL spoken document retrieval system is based on the ABBOT large vocabulary continuous speech recognizer [1] and a probabilistic ranked text retrieval system.  ... 
doi:10.1109/icassp.1998.679707 dblp:conf/icassp/AbberleyRC98 fatcat:2o2ohirf5jboxnucy7lwcklk3u

The THISL SDR system at TREC-8

Dave Abberley, Steve Renals, Daniel P. W. Ellis, Tony Robinson
2017
This paper describes the participation of the THISL group at the TREC-8 Spoken Document Retrieval (SDR) track.  ...  The THISL SDR system consists of the realtime version of the ABBOT large vocabulary speech recognition system and the THISLIR text retrieval system.  ...  The THISL 1 spoken document retrieval system consists of the 'real time' version of the ABBOT large vocabulary continuous speech recognizer [2] and the THISLIR text retrieval system [3] .  ... 
doi:10.7916/d8fj2s36 fatcat:crz5utpn6nhq3g7gtc33dyzkga

European projects update

Daniel P. W. Ellis
2017
Report on European meetings of the Thematic Indexing of Spoken Language (THISL) and Recognition of Speech by Partial Information Techniques (RESPITE) projects.  ...  project: Using ASR (&c) to index BBC news archives • ESCA workshop on Spoken Document Retrieval (SDR) -April, Cambridge -systems, IR/IE -demos, including thislIR • Current actions: -finalize  ...  Thisl final year planning meeting • Use artificial mixtures to train R xx → SNR map • 'Full combination' multistream needs weights:-p( q | a,b,c,d ) = ∑ S p( S ) . p( q | S,a,b,c,d )S ranges over 16 possible  ... 
doi:10.7916/d87p96nh fatcat:2y46frmryjgb5fxlr5mkb4bawq

Web-assisted annotation, semantic indexing and search of television and radio news

Mike Dowman, Valentin Tablan, Hamish Cunningham, Borislav Popov
2005 Proceedings of the 14th international conference on World Wide Web - WWW '05  
The Rich News system, that can automatically annotate radio and television news with the aid of resources retrieved from the World Wide Web, is described.  ...  The text and meta-data of the web pages is then used to create index documents for the stories in the original broadcasts, which are semantically annotated using the KIM knowledge management platform.  ...  ACKNOWLEDGMENTS The research for this paper was conducted as part of the European Union Sixth Framework Program projects PrestoSpace (FP6-507336) and SEKT (EU IST IP 2003-506826).  ... 
doi:10.1145/1060745.1060781 dblp:conf/www/DowmanTCP05 fatcat:s3u5pphtundlhd7towk6b5uxbi

An overview of Speech Recognition research at ICSI

Daniel P. W. Ellis
2017
Overview of speech recognition research at the International Computer Science Institute, and introduction to connectionist speech recognition.  ...  • Information Retrieval (IR) -TREC/MUC 'spoken documents' -tolerant of word error rate, e.g.: F0: THE VERY EARLY RETURNS OF THE NICARAGUAN PRESIDENTIAL ELECTION SEEMED TO FADE BEFORE THE LOCAL  ...  WER ratio plp + msg Feature combo 74.1% plp + msg Prob. combo 63.0% plp + msg HTK on probs. 51.6% .2 Spoken document retrieval • Based on DARPA/NIST Broadcast News • Training material  ... 
doi:10.7916/d8n01frj fatcat:rh7ztx52wfdglkr67rsrcenmle

A unified language model for large vocabulary continuous speech recognition of Turkish

Ebru Arısoy, Helin Dutağacı, Levent M. Arslan
2006 Signal Processing  
The proposed model resulted in letter error rates (LER's) of approximately 28% for a speaker independent system and 20% for a speaker dependent system.  ...  A combined model is proposed which aims to produce a balance between the OOV rate and the amount of phoneme sequence constraints on recognition units.  ...  Acknowledgements The authors would like to thank Dr. Murat Saraclar for the discussions and Hacettepe University Radiology Department for their help in supplying radiological reports.  ... 
doi:10.1016/j.sigpro.2005.12.002 fatcat:kvh2k6pr5feg3nkoenuxnnifra

Cohort Profile: the Health and Retirement Study (HRS)

A. Sonnega, J. D. Faul, M. B. Ofstedal, K. M. Langa, J. W. Phillips, D. R. Weir
2014 International Journal of Epidemiology  
The HRS has been a leading force for rapid release of data while simultaneously protecting the confidentiality of respondents.  ...  The Health and Retirement Study (HRS) is a nationally representative longitudinal survey of more than 37 000 individuals over age 50 in 23 000 households in the USA.  ...  Acknowledgments HRS gratefully acknowledges the contribution of the study participants who have given countless hours of their time to make this study what it is. Conflicts of interest: None declared.  ... 
doi:10.1093/ije/dyu067 pmid:24671021 pmcid:PMC3997380 fatcat:7jb323gp5jg7xpddbgkuwyb4xu

Speech Recognition at ICSI: Broadcast News and beyond

Daniel P. W. Ellis
2017
Overview of the International Computer Science Institute's efforts, in collaboration with European partners, to prepare a system to submit to a Defense Advanced Research Projects Agency and National Institute  ...  • Information Retrieval (IR) -TREC/MUC 'spoken documents' -tolerant of word error rate, e.g.: F0: THE VERY EARLY RETURNS OF THE NICARAGUAN PRESIDENTIAL ELECTION SEEMED TO FADE BEFORE THE LOCAL  ...  Indexing of Spoken Language (Thisl) • EC collaboration, BBC providing data • > 500 hr archive data • IR is key factor -stop lists -weighting schemes -query expansion Control Text Audio  ... 
doi:10.7916/d8qr55c1 fatcat:puihxcd2o5f33o2dgncyz4schm
« Previous Showing results 1 — 15 out of 26 results