this paper, we study alternatives to perplexity for predicting language model performance, including other global features as well as a new approach that predicts, with a high correlation (0.96), performance differences associated with localized changes in language models given a recognition system. Experiments focus on the problem of augmenting in-domain Switchboard text with out-of-domain text from Wall Street Journal and Broadcast News that differ in both style and content from the in-domain data.

Abstract. The deployment and integration of audio processing tools can enhance the semantic annotation of multimedia content, and as a consequence, improve the effectiveness of conceptual access tools. This paper overviews the various ways in which automatic speech and audio analysis can contribute to increased granularity of automatically extracted metadata. A number of techniques will be presented, including the alignment of speech and text resources, large vocabulary speech recognition, key word spotting and speaker classification. The applicability of techniques will be discussed from a media crossing perspective. The added value of the techniques and their potential contribution to the content value chain will be illustrated by the description of two (complementary) demonstrators for browsing broadcast news archives. 1

Statistical language models encode linguistic information in such a way as to be useful to systems which process human language. Such systems include those for optical character recognition and machine translation. Currently, however, the most common application of language modelling is in automatic speech recognition, and it is this that forms the focus of this thesis. Most current speech recognition systems are dedicated to one specific task (for example, the recognition of broadcast news), and thus use a language model which has been trained on text which is appropriate to that task. If, however, one wants to perform recognition on more general language, then creating an appropriate language model is far from straightforward. A taskspecific language model will often perform very badly on language from a different domain, whereas a model trained on text from many diverse styles of language might perform better in general, but will not be especially well suited to any particular domai...

Speech recognition problems are a reality in current spoken dialogue systems. In order to better understand these phenomena, we study dependencies between speech recognition problems and several higher level dialogue factors that define our notion of student state: frustration/anger, certainty and correctness. We apply Chi Square (χ2) analysis to a corpus of speech-based computer tutoring dialogues to discover these dependencies both within and across turns. Significant dependencies are combined to produce interesting insights regarding speech recognition problems and to propose new strategies for handling these problems. We also find that tutoring, as a new domain for speech applications, exhibits interesting tradeoffs and new factors to consider for spoken dialogue design. 1

In this paper we investigate measures for the evaluation of pronunciation dictionaries that can be used independently of the type of lexicon, the language, a specific recognizer and how the dictionary was generated. We will describe statistical measures, measures based on information theory and performance measures and give examples how these measures can be practically applied in supervision of data-driven dictionary training, selection of pronunciation variants and evaluation of the consistency of different dictionaries. Although the introduced measures are independent of the type of dictionary, we only report results obtained with a datadriven dictionary generation and do not address measures specific to rule-based approaches. 1.

This thesis addresses the application of automatic speech recognition to the task of offline closed-captioning of television programs, and describes the collection of corpora to support such research and an exploration of issues to be addressed. The use of automatic speech recognition (ASR) for transcription of broadcast speech and as an aid to captioning is reviewed. As background to the task, the methodology for large vocabulary continuous speech recognition (LVCSR) is presented, with particular attention given to the issues of large vocabulary language modelling and consideration of the acoustic complexity arising in broadcast material. A speech corpus of segmented and transcribed speech utterances for ten program episodes was developed for a typical genre of television programming (travelogues) for which offline closed-captions are applied. The development of this corpus demonstrates the feasibility of using existing closed-caption sources for generating labelled acoustic data suitable for speech recognition research. The speech corpus exhibits far greater acoustic complexity and much lower signal to noise ratios than occurs in broadcast news data (which has been systematically evaluated in ASR research). Noise-tolerant speech recognisers were developed and effectively

Abstract not found

Decoders that make use of token-passing restrict their search space by various types of token pruning. With use of the Language Model Look-Ahead (LMLA) technique it is possible to increase the number of tokens that can be pruned without loss of decoding precision. Unfortunately, for token passing decoders that use single static pronunciation prefix trees, full n-gram LMLA increases the needed number of language model probability calculations considerably. In this paper a method for applying full n-gram LMLA in a decoder with a single static pronunciation tree is introduced. The experiments show that this method improves the speed of the decoder without an increase of search errors.