G.J.F. Jones, J.T. Foote, K. Sparck-Jones and S.J. Young, "Retrieving spoken documents by combining multiple index sources", Proc. SIGIR'96, Zurich, August 1996, pp: 30-39.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... methods have been explored in the information retrieval community, which can be classified into the following two fundamental categories: spoken document retrieval, in which written queries are used to search speech (e.g. broadcast news audio) archives for relevant speech information [5, 6, 15 17, 19, 20], speech driven (spoken query) retrieval, in which spoken queries are used to retrieve relevant textual information [2, 3] Initiated partially by the TREC 6 spoken document retrieval (SDR) track [4] various methods have been proposed for spoken document retrieval. However, a relatively ....

G. Jones, J. Foote, K. S. Jones, and S. Young. Retrieving spoken documents by combining multiple index sources. In Proceedings of the 19th Annual International pages 30--38, 1996.

....A word based system however can use an index with a relatively constant access time regardless of size. This search problem can be alleviated by building an index of sequences phonemes or syllables (e.g. 5] Approaches which combine word and phoneme models have also been tried (e.g. [7], 8] Typically, linear combinations are considered. The theoretical properties of linearly combined indexes are studied in [9] Here it is noted that the usefulness of linear combination is limited to certain situations. The main problem is that it is not known how to optimize the combination ....

G. J. F. Jones, J. T. Foote, K. Spark Jones, and S. J. Young, "Retrieving spoken documents by combining multiple index sources," in SIGIR1996.

.... methods have been explored in the information retrieval (IR) community, which can be classified into the following two fundamental categories: spoken document retrieval, in which written queries are used to search speech (e.g. broadcast news audio) archives for relevant speech information [10, 11, 18, 19, 20, 22, 23], speech driven retrieval, in which spoken queries are used to retrieve relevant textual information [2, 3, 4, 5, 7, 14] Initiated partially by the TREC 6 spoken document retrieval (SDR) track [6] various methods have been proposed for spoken document retrieval. However, a relatively small ....

G. Jones, J. Foote, K. S. Jones, and S. Young. Retrieving spoken documents by combining multiple index sources. In Proceedings of the 19th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 30--38, 1996.

....containing n grams of di#erent lengths. Various combinations were tried, varying from combination of two ngram lengths to combining all n gram lengths. By combining n grams of di#erent length, we can investigate whether di#erent n gram sizes provide di#erent forms of evidence for retrieval [9]. Combination permitted us to investigate whether boundary information and stopping a#ected retrieval on larger collections. When phoneme recognition was performed on the training collection of TREC 6 SDR, we found that, for each recognised word, similar recognition errors were occurring ....

G. J. F. Jones, J. T. Foote, K. Sparck Jones, and S. J. Young. Retrieving spoken documents by combining multiple index sources. In Proceedings of the Nineteenth ACM-SIGIR Conference on Research and Development in Information Retrieval, pages 30 -- 38, 1996.

.... group and by others, has shown that spoken document re trieval using speech recognition is becoming practical [12, 9] Future work in this project will be to integrate different audio index sources available from large vocabulary recognition and conventional as well as lattice based word spotting [14, 13]. In addition, work will need to be done to make the system robust to environmental noise, microphone differences, accent variability, and telephone bandwidth speech. Another promising area is to use other types of audio information, such as speaker or music identification, to help index ....

G. J. F. Jones, J. T. Foote, K. Sp/irck Jones, and S. J. Young. Retrieving spoken documents by combining multiple index sources. In Proc. SIGIR 96, Ziirich, August 1996. ACM.

....they still require significant search effort. In the future, we expect to combine phone lattice based spotting with large vocabulary transcription and our preliminary experiments in this area have shown that this combination can give better retrieval per formance than either technique alone [8]. In a similar fashion to the CMU Informedia project [9] we have also developed a version of the VMR system which uses dose caption text transmissions to index broadcast news material [10] This allows a user to browse rapidly through several months of broadcast news looking for items of ....

G. J. F. Jones, J. T. Foote, K. Spirck Jones, and S. J. Young. Retrieving spoken documents by combining multiple index sources. In Proc. SIGIR 96, Z'drich, August 1996. ACM.

....regions [2] The goal of the current research is to provide UIs to support both search and visual scanning of speech data. Previous approaches to speech access have employed three main techniques. The first uses different types of structural indices to access different speech regions: by speaker [5,7,10,21], emphasis [1,15] external events such as user note taking behaviors [7,11,15,20,jm in here ] or accompanying visual events [3,6] Indices are then represented visually, allowing browsing and random access to relevant regions. The second approach involves contentbased search. Automatic speech ....

....[3,6] Indices are then represented visually, allowing browsing and random access to relevant regions. The second approach involves contentbased search. Automatic speech recognition (ASR) is applied to speech, and the resulting errorful text is then searched using information retrieval techniques [9,10]. The third is surface manipulation signal processing techniques are applied to digital speech allowing it to be played back at several times its normal rate, retaining comprehensibility [1] Our own research combines all three techniques in order to access speech data. We chose voicemail as ....

[Article contains additional citation context not shown here]

Jones, G., Foote, J., Sprck Jones, K., and Young, S. Retrieving Spoken Documents by Combining Multiple Index Sources, In SIGIR96, 30-38, 1996.

....has yet to be tested. A promising approach to the integration of LVCSR and IR, and one that will potentially enhance system resistance to recognizer noise, is the combination of multimodal information sources, such a system combines the output of LVCSR systems based on different sub word units [11] [24] ....

G. J. F. Jones, J.T. Foote, K. Sparck Jones and S.J. Young. Retrieving Spoken Documents by Combining Multiple Index Sources, SIGIR 1996.

....documents, images, audio and video recordings. DART aims to provide the means to index, annotate, navigate and retrieve from diverse collections of these assets. The project was motivated in part by our successful collaboration with Cambridge University in the Video Mail Retrieval project [2,3] but it has evolved to encompass a much broader range of issues in multimedia information retrieval. In developing the core DART technologies, we have built an application to help manage personal photograph collections. With the increasing popularity of digital cameras, the cost of producing ....

....0.6788 Table 1: Retrieval precision. On the whole, annotations have proved to be a useful means of retrieving photographs, despite the inaccuracies introduced by speech recognition. Degradation in retrieval performance due to speech recognition errors is roughly comparable to that reported in [2]. However, during our experiments, some problems were evident. While roll annotations were often used to label related groups of images (e.g. These photos were taken while on holiday in Poland ) they were sometimes misused. In particular, for those photographs which had been scanned from APS ....

G. J. F. Jones, J. T. Foote, K. Sparck Jones and S. J. Young.. Retrieving Spoken Documents by Combining Multiple Index Sources. Proceedings of the ACM SIGIR Conference, Zurich, Switzerland, 1996.

.... 3 Related Work Some early work on word based speech processing that resembles information retrieval methods was done by Rose et al. using word spotting [20] More extensive work on wordspotting based speech retrieval was done in the Video Mail Retrieval (VMR) project at Cambridge University [11]. An alternative to word based approaches is to recognize sub word units (for instance, phones) and use sequences of these sub word units as index terms [29] However, it is unclear if the results from this approach are competitive with word based approaches now that very large vocabulary ....

....it is unclear if the results from this approach are competitive with word based approaches now that very large vocabulary recognition systems are available. It is also possible to use simultaneously as index terms words from the best word transcription and phonetic n grams from phone lattices [11, 31]. A comparison of the indexing effectiveness of various sub word units is done in [15] Although we experiment only with word based systems in the present study, the techniques explored here are quite general. The methods we propose would enhance the index representation for speech documents ....

G.J.F. Jones, J.T. Foote, K. Sparck Jones, and S.J. Young. Retrieving spoken documents by combining multiple index sources. In Proceedings of SIGIR'96, pages 30--38. ACM, New York, August 1996.

.... years, various systems have been built to enable capture and browsing of spoken conversational data from meetings and recorded lectures [6, 9, 10, 17, 15] and personally dictated information [2, 13] Other systems allow search of multimedia archives of television programmes [5, 11] and videomail [8]. While extensive evaluations of this technology remain to be carried out, naturalistic studies of audio browsing systems demonstrate their effectiveness in helping users produce accurate meeting summaries [10, 15, 16] These and other studies also showed that indexed audio produces more accurate ....

G. J. F. Jones, J. T. Foote, K. S. Jones, and S. J. Young. Retrieving spoken documents by combining multiple index sources. In Proceedings of SIGIR 96, Zurich, August 1996. ACM.

....on full word recognition techniques while others have been based on indexing spoken text by the phones. These approaches can be roughly grouped as follows. Approach 1: Word Spotting Instead of trying to do recognition of full speech, the Cambridge Olivetti VMR (video mail retrieval) project [6] did word spotting, i.e. given a pre defined vocabulary of the order of some tens of words, process the spoken audio component of video mail looking for these words and these words only, and use them as indexing terms. By this restriction to a reduced set of key words the problem of speech ....

Jones, G.J.F., Foote, J.T., Sprck Jones, K. and Young, S.J.: Retrieving spoken documents by combining multiple index sources. In Proceedings of SIGIR 96, Research and Development in Information Retrieval, 30-38, Zrich, ACM Press, (1996).

.... use phone sequences (phone n grams) generated by post processing the output of a phonetic speech recognizer [31,49] There are also methods that search for the query terms on phonetic transcriptions or phone lattice representations of the speech messages instead of creating subword indexing terms [9,23,26,48]. Some of these methods combine subword and large vocabulary word approaches to address the issue of new words in the queries [23,26] However, there have been few studies that explore the space of possible subword unit representations to determine the complexity of the subword units needed to ....

.... that search for the query terms on phonetic transcriptions or phone lattice representations of the speech messages instead of creating subword indexing terms [9,23,26,48] Some of these methods combine subword and large vocabulary word approaches to address the issue of new words in the queries [23,26]. However, there have been few studies that explore the space of possible subword unit representations to determine the complexity of the subword units needed to perform e#ective spoken document retrieval and to measure the behavior and sensitivity of di#erent types of subword units to speech ....

G. J. F. Jones, J. T. Foote, K. S. Jones, and S. J. Young, "Retrieving spoken documents by combining multiple index sources," in Proceedings of the Nineteenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, Zurich, Switzerland, pp. 30--39, Aug. 1996.

.... Figure 1: Overview of the spoken document system architecture Previous work on information retrieval from speech databases include a system for Swiss radio news [14] a system for a digital video library [15] radio news broadcast retrieval using subword units [9] a Video Mail Retrieval system [4,5] and a number of systems developed for the Text Retrieval Conference Spoken Document Retrieval track [13] inter alia. 2. SPEECH RECOGNITION The speech recognition component of SCAN includes an intonational phrase boundary detection module and a classification module. These subcomponents ....

Jones, G.J.F., Foote, J.T., Sparck-Jones, K., and Young, S.J., "Retrieving spoken documents by combining multiple index sources", Proceedings of the Nineteenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, 1996, pp. 30-38.

.... have been successful for accessing personal speech data [15,19,24] Several video retrieval systems have presented key video frames to provide visual overviews to video programs [7,18] Other broadcast news and meeting recording systems present high level topic or speaker switching information [10,11,16]. However, with the exception of [11] these latter UIs have not been evaluated on access tasks with real users. Figure 4: Effects of browser and task difficulty on perceived task difficulty We find it significant in our studies that the multimodal SCAN interface is beneficial only for certain ....

Jones, G., Foote, J. T., Sparck Jones, K. and Young, S. J. Retrieving spoken documents by combining multiple index sources. In Proceedings of SIGIR-96, Zurich, 1996.

....supporting research on information retrieval strategies for machine generated speech transcripts, the system is also a testbed for user interface experiments on intelligent presentation of speech documents to users. Previous work on spoken document retrieval includes a Video Mail Retrieval system [7,8], radio news broadcast retrieval using subword units [16] a retrieval system for a digital video library [25] a system for Swiss radio news [24] and the systems developed for the TREC 6 SDR track [23] inter alia. 2. SYSTEM OVERVIEW An overview of the system architecture is provided in Figure ....

....3 and 4, retrieval from automatic transcriptions is about 70 80 as effective as retrieval from human transcriptions. This number is in agreement with other studies that have compared the retrieval effectiveness on both automatically generated transcriptions and human generated transcriptions [8,25]. The second conclusion we draw is simply that better recognition contributes to better retrieval. While this seems intuitively transparent, we are pleased that our empirical findings support this hypothesis. The relatively small size of the HUB4 corpus for information retrieval purposes initially ....

Jones, G. J. F., Foote, J. T., Sparck-Jones, K., and Young, S. J., "Retrieving spoken documents by combining multiple index sources", Proceedings of the Nineteenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, 1996, pp. 3038.

No context found.

VMR: G.J.F. Jones, J.T. Foote, K. Sparck Jones and S.J. Young, `Retrieving spoken documents by combining multiple index sources', SIGIR-96, Proceedings of the 19th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, 1996, 30-38.

No context found.

G. J. F. Jones, J. T. Foote, K. Sp#arck Jones, and S. J. Young. Retrieving spoken documents by combining multiple index sources. In Proc. SIGIR 96,Z#urich, August 1996. ACM.

No context found.

G.J.F. Jones, J.T. Foote, K. Sparck-Jones and S.J. Young, "Retrieving spoken documents by combining multiple index sources", Proc. SIGIR'96, Zurich, August 1996, pp: 30-39.

No context found.

Jones, Gareth J.F., J.T.Foote, K. Sparck-Jones and S. Young, Retrieving Spoken Documents by Combining Multiple Index Sources, Proceedings of ACM-SIGIR 1996, Zrich.

No context found.

G. J. F. Jones, J. T. Foote, K. Spark Jones, and S. J. Young. Retrieving spoken documents by combining multiple index sources. In SIGIR1996.

No context found.

Jones, G.J.F., Foote, J.T., Sprck Jones, K., and Young, S.J., "Retrieving Spoken Documents by Combining Multiple Index Sources", SIGIR-96 Proceedings of the 1996 ACM SIGIR Conference, Zurich, Switzerland.

No context found.

Jones, G., Foote, J., Sprck Jones, K., and Young, S. Retrieving Spoken Documents by Combining Multiple Index Sources, In SIGIR96, 30-38, 1996.

No context found.

G.J.F. Jones et al., "Retrieving Spoken Documents by Combining Multiple Index Sources," Proc. SIGIR 96, ACM Press, New York, 1996, pp. 30-38.

No context found.

Jones, G.; Foote, J.; Sparck Jones, K.; Young, S. (1996). Retrieving Spoken Documents by Combining Multiple Index Sources. In [Frei et al. 96], pages 30--39.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC