Text, Speech and Language Technology – serie

Corpus-based methods will be found at the heart of many language and speech processing systems. This book provides an in-depth introduction to these technologies through chapters describing basic statistical modeling techniques for language and speech, the use of Hidden Markov Models in continuous speech recognition, the development of dialogue systems, part-of-speech tagging and partial parsing, data-oriented parsing and n-gram language modeling. The book attempts to give both a clear overview of the main technologies used in language and speech processing, along with sufficient mathematics to understand the underlying principles. There is also an extensive bibliography to enable topics of interest to be pursued further. Overall, we believe that the book will give newcomers a solid introduction to the field and it will give existing practitioners a concise review of the principal technologies used in state-of-the-art language and speech processing systems. Corpus-Based Methods in Language and Speech Processing is an initiative of ELSNET, the European Network in Language and Speech. In its activities, ELSNET attaches great importance to the integration of language and speech, both in research and in education. The need for and the potential of this integration are well demonstrated by this publication.

An Introduction to Text-to-Speech Synthesis is a comprehensive introduction to the subject. The author treats two areas of speech synthesis: Part I of the book concerns natural language processing and the inherent problems it presents for speech synthesis; Part II focuses on digital signal processing, with an emphasis on the concatenative approach. Both parts of the text guide the reader through the material in a step-by-step easy-to-follow way. This is the first book to treat the topic of speech synthesis from the perspective of two different engineering approaches. The book will be of interest to researchers and students in phonetics and speech communication, in both academia and industry.

Researchers in a number of disciplines deal with large text sets requiring both text management and text analysis. Faced with a large amount of textual data collected in marketing surveys, literary investigations, historical archives and documentary data bases, these researchers require assistance with organizing, describing and comparing texts. Exploring Textual Data demonstrates how exploratory multivariate statistical methods such as correspondence analysis and cluster analysis can be used to help investigate, assimilate and evaluate textual data. The main text does not contain any strictly mathematical demonstrations, making it accessible to a large audience. This book is very user-friendly with proofs abstracted in the appendices. Full definitions of concepts, implementations of procedures and rules for reading and interpreting results are fully explored. A succession of examples is intended to allow the reader to appreciate the variety of actual and potential applications and the complementary processing methods. A glossary of terms is provided.

This book is a revised version of my doctoral thesis which was submitted in April 1993. The main extension is a chapter on evaluation of the system de­ scribed in Chapter 8 as this is clearly an issue which was not treated in the original version. This required the collection of data, the development of a concept for diagnostic evaluation of linguistic word recognition systems and, of course, the actual evaluation of the system itself. The revisions made primarily concern the presentation of the latest version of the SILPA system described in an additional Subsection 8. 3, the development environment for SILPA in Sec­ tion 8. 4, the diagnostic evaluation of the system as an additional Chapter 9. Some updates are included in the discussion of phonology and computation in Chapter 2 and finite state techniques in computational phonology in Chapter 3. The thesis was designed primarily as a contribution to the area of compu­ tational phonology. However, it addresses issues which are relevant within the disciplines of general linguistics, computational linguistics and, in particular, speech technology, in providing a detailed declarative, computationally inter­ preted linguistic model for application in spoken language processing. Time Map Phonology is a novel, constraint-based approach based on a two-stage temporal interpretation of phonological categories as events.

This volume is a selection of papers presented at a workshop entitled Predicative Forms in Natural Language and in Lexical Knowledge Bases organized in Toulouse in August 1996. A predicate is a named relation that exists among one or more arguments. In natural language, predicates are realized as verbs, prepositions, nouns and adjectives, to cite the most frequent ones. Research on the identification, organization, and semantic representa­ tion of predicates in artificial intelligence and in language processing is a very active research field. The emergence of new paradigms in theoretical language processing, the definition of new problems and the important evol­ ution of applications have, in fact, stimulated much interest and debate on the role and nature of predicates in naturallangage. From a broad theoret­ ical perspective, the notion of predicate is central to research on the syntax­ semantics interface, the generative lexicon, the definition of ontology-based semantic representations, and the formation of verb semantic classes. From a computational perspective, the notion of predicate plays a cent­ ral role in a number of applications including the design of lexical knowledge bases, the development of automatic indexing systems for the extraction of structured semantic representations, and the creation of interlingual forms in machine translation.

The last decade has been one of dramatic progress in the field of Natural Language Processing (NLP). This hitherto largely academic discipline has found itself at the center of an information revolution ushered in by the Internet age, as demand for human-computer communication and informa­ tion access has exploded. Emerging applications in computer-assisted infor­ mation production and dissemination, automated understanding of news, understanding of spoken language, and processing of foreign languages have given impetus to research that resulted in a new generation of robust tools, systems, and commercial products. Well-positioned government research funding, particularly in the U. S. , has helped to advance the state-of-the­ art at an unprecedented pace, in no small measure thanks to the rigorous 1 evaluations. This volume focuses on the use of Natural Language Processing in In­ formation Retrieval (IR), an area of science and technology that deals with cataloging, categorization, classification, and search of large amounts of information, particularly in textual form. An outcome of an information retrieval process is usually a set of documents containing information on a given topic, and may consist of newspaper-like articles, memos, reports of any kind, entire books, as well as annotated image and sound files. Since we assume that the information is primarily encoded as text, IR is also a natural language processing problem: in order to decide if a document is relevant to a given information need, one needs to be able to understand its content.

Techniques in Speech Acoustics provides an introduction to the acoustic analysis and characteristics of speech sounds. The first part of the book covers aspects of the source-filter decomposition of speech, spectrographic analysis, the acoustic theory of speech production and acoustic phonetic cues. The second part is based on computational techniques for analysing the acoustic speech signal including digital time and frequency analyses, formant synthesis, and the linear predictive coding of speech. There is also an introductory chapter on the classification of acoustic speech signals which is relevant to aspects of automatic speech and talker recognition. The book intended for use as teaching materials on undergraduate and postgraduate speech acoustics and experimental phonetics courses; also aimed at researchers from phonetics, linguistics, computer science, psychology and engineering who wish to gain an understanding of the basis of speech acoustics and its application to fields such as speech synthesis and automatic speech recognition.

This introduction to the acoustic analysis and characteristics of speech sounds covers aspects of the source-filter decomposition of speech, spectrographic analysis, the acoustic theory of speech production and acoustic phonetic cues. The second part is based on computational techniques for analysing the acoustic speech signal including digital time and frequency analyses, formant synthesis, and the linear predictive coding of speech. There is also an introductory chapter on the classification of acoustic speech signals which is relevant to aspects of automatic speech and talker recognition. Included with the book is a CD-ROM containing extensive speech corpora, the EMU speech analysis tools, extensions to the X-LISP-STAT programming language that are adapted to speech analysis, and numerous exercises that are linked to the major themes of the book and which can be run on Windows-95 and UNIX platforms.The book and CD-ROM are intended for use as teaching materials on undergraduate and postgraduate speech acoustics and experimental phonetics courses; they are also aimed at researchers from phonetics, linguistics, computer science, psychology and engineering who wish to gain an understanding of the basis of speech acoustics and its application to fields such as speech synthesis and automatic speech recognition.

In both the linguistic and the language engineering community, the creation and use of annotated text collections (or annotated corpora) is currently a hot topic. Annotated texts are of interest for research as well as for the development of natural language pro­ cessing (NLP) applications. Unfortunately, the annotation of text material, especially more interesting linguistic annotation, is as yet a difficult task and can entail a substan­ tial amount of human involvement. Allover the world, work is being done to replace as much as possible of this human effort by computer processing. At the frontier of what can already be done (mostly) automatically we find syntactic wordclass tagging, the annotation of the individual words in a text with an indication of their morpho syntactic classification. This book describes the state of the art in syntactic wordclass tagging. As an attempt to give an overall view of the field, this book is of interest to (at least) two, possibly very different, types of reader. The first type consists of those people who are using, or are planning to use, tagged material and taggers. They will want to know what the possibilities and impossibilities of tagging are, but are not necessarily interested in the internal working of automatic taggers. This, on the other hand, is the main interest of our second type of reader, the builders of automatic taggers and other natural language processing software.

Most books about computational (lexical) semantic lexicons deal with the depth (or content) aspect of lexicons, ignoring the breadth (or coverage) aspect. This book presents an attempt to address both issues: content and coverage of computational semantic lexicons. Moreover, it addresses issues in implemented systems such as the application time of lexical rules. Lexical rules and lexical underspecification are also contrasted in implemented systems. The main approaches in the field of computational (lexical) semantics are represented in the present book (including Wordnet, CyC, Mikrokosmos, Generative Lexicon). This book embraces several fields (and subfields) as different as: linguistics (theoretical, computational, semantics, pragmatics), psycholinguistics, cognitive science, computer science, artificial intelligence, knowledge representation, statistics and natural language processing.

ABOUT THIS BOOK This book is intended for researchers who want to keep abreast of cur­ rent developments in corpus-based natural language processing. It is not meant as an introduction to this field; for readers who need one, several entry-level texts are available, including those of (Church and Mercer, 1993; Charniak, 1993; Jelinek, 1997). This book captures the essence of a series of highly successful work­ shops held in the last few years. The response in 1993 to the initial Workshop on Very Large Corpora (Columbus, Ohio) was so enthusias­ tic that we were encouraged to make it an annual event. The following year, we staged the Second Workshop on Very Large Corpora in Ky­ oto. As a way of managing these annual workshops, we then decided to register a special interest group called SIGDAT with the Association for Computational Linguistics. The demand for international forums on corpus-based NLP has been expanding so rapidly that in 1995 SIGDAT was led to organize not only the Third Workshop on Very Large Corpora (Cambridge, Mass. ) but also a complementary workshop entitled From Texts to Tags (Dublin). Obviously, the success of these workshops was in some measure a re­ flection of the growing popularity of corpus-based methods in the NLP community. But first and foremost, it was due to the fact that the work­ shops attracted so many high-quality papers.

This book offers a state-of-the-art survey of methods and techniques for structuring, acquiring and maintaining lexical resources for speech and language processing. The first chapter provides a broad survey of the field of computational lexicography, introducing most of the issues, terms and topics which are addressed in more detail in the rest of the book. The next two chapters focus on the structure and the content of man-made lexicons, concentrating respectively on (morpho-)syntactic and (morpho-)phonological information. Both chapters adopt a declarative constraint-based methodology and pay ample attention to the various ways in which lexical generalizations can be formalized and exploited to enhance the consistency and to reduce the redundancy of lexicons. A complementary perspective is offered in the next two chapters, which present techniques for automatically deriving lexical resources from text corpora. These chapters adopt an inductive data-oriented methodology and focus also on methods for tokenization, lemmatization and shallow parsing. The next three chapters focus on speech applications, more specifically on the organization of speech data bases, and on the use of lexica in speech synthesis and speech recognition. The last chapter takes a psycholinguistic perspective and addresses the relation between storage and computation in the mental lexicon. The relevance of these topics for speech and language processing is obvious, for since NLP systems need large lexica in order to achieve reasonable coverage, and since the construction and maintenance of large-size lexical resources is a complex and costly task, it is of crucial importance for those who design or build such systems to be aware of the latest developments in this fast-moving field. The intended audience for this book includes advanced students and professional scientists working in the areas of computational linguistics and language and speech technology.

This work offers a survey of methods and techniques for structuring, acquiring and maintaining lexical resources for speech and language processing. The first chapter provides a broad survey of the field of computational lexicography, introducing most of the issues, terms and topics which are addressed in more detail in the rest of the book. The next two chapters focus on the structure and the content of man-made lexicons, concentrating respectively on (morpho-)syntactic and (morpho-)phonological information. Both chapters adopt a declarative constraint-based methodology and pay ample attention to the various ways in which lexical generalizations can be formalized and exploited to enhance the consistency and to reduce the redundancy of lexicons. A complementary perspective is offered in the next two chapters, which present techniques for automatically deriving lexical resources from text corpora. These chapters adopt an inductive data-oriented methodology and focus also on methods for tokenization, lemmatization and shallow parsing. The next three chapters focus on speech synthesis and speech recognition.The last chapter takes a psycholinguistic perspective and addresses the relation between storage and computation in the mental lexicon. The work should be helpful for advanced students and professional scientists working in the areas of computational linguistics and language and speech technology.

l This book evolved from the ARCADE evaluation exercise that started in 1995. The project's goal is to evaluate alignment systems for parallel texts, i. e. , texts accompanied by their translation. Thirteen teams from various places around the world have participated so far and for the first time, some ten to fifteen years after the first alignment techniques were designed, the community has been able to get a clear picture of the behaviour of alignment systems. Several chapters in this book describe the details of competing systems, and the last chapter is devoted to the description of the evaluation protocol and results. The remaining chapters were especially commissioned from researchers who have been major figures in the field in recent years, in an attempt to address a wide range of topics that describe the state of the art in parallel text processing and use. As I recalled in the introduction, the Rosetta stone won eternal fame as the prototype of parallel texts, but such texts are probably almost as old as the invention of writing. Nowadays, parallel texts are electronic, and they are be­ coming an increasingly important resource for building the natural language processing tools needed in the "multilingual information society" that is cur­ rently emerging at an incredible speed. Applications are numerous, and they are expanding every day: multilingual lexicography and terminology, machine and human translation, cross-language information retrieval, language learning, etc.

The study of prosody is perhaps the area of speech research which has undergone the most noticeable development during the past ten to fifteen years. As an indication of this, one can note, for example, that at the latest International Conference on Spoken Language Processing in Philadelphia (October 1996), there were more sessions devoted to prosody than to any other area. Not only that, but within other sessions, in particular those dealing with dialogue, several of the presentations dealt specifically with prosodic aspects of dialogue research. Even at the latest Eurospeech meeting in Rhodes (September 1997), prosody, together with speech recognition (where several contributions dealt with how prosodic cues can be exploited to improve recognition processes) were the most frequent session topics, despite the fact that th'ere was a separate ESCA satellite workshop on intonation in conjunction with the main Eurospeech meeting which included over 80 contributions. This focus on prosodic research is partly due to the fact that developments in speech technology have made it possible to examine the acoustic parameters associated with prosodic phenomena (in particular fundamental frequency and duration) to an extent which has not been possible in other domains of speech research. It is also due to the fact that significant theoretical advances in linguistics and phonetics have been made during this time which have made it possible to obtain a better understanding of how prosodic parameters function in expressing different kinds of meaning in the languages of the world.

ANTONIS BOTINIS 1. 1 Background This introduction provides essential information about the structure and the objects of study of this volume. Following the introduction, fourteen papers which represent current research on intonation are organised into five thematic sections: (I) Overview of Intonation, (II) Prominence and Focus, (III) Boundaries and Discourse, (IV) Intonation Modelling, and (V) Intonation Technology. Within the sections the papers are arranged thematically, although several papers which deal with various aspects of intonation and prosody are basically intersectional. As the title indicates, "Intonation: Analysis, Modelling and Technology" is a contribution to the study of prosody, with major emphasis on intonation. Intonation and tonal themes are thus the central object of the volume, although temporal and dynamic aspects are also taken into consideration by a good number of papers. Although tonal and prosodic distinctions have been dealt with throughout man's literate history with reference to the study of language, for example by classical philosophers such as Plato and Aristotle, it is in recent decades that we have witnessed the most fertile growth in intonation studies, as with experimental phonetics and speech technology in general. As Rossi (this volume) points out, intonation research really began to blossom in the sixties with a multi­ fold increase in prosodic studies, reflected in contributions to the International Congress of Phonetic Sciences (ICPhS), and in the international literature.

Parsing technology is concerned with finding syntactic structure in language. In parsing we have to deal with incomplete and not necessarily accurate formal descriptions of natural languages. Robustness and efficiency are among the main issuesin parsing. Corpora can be used to obtain frequency information about language use. This allows probabilistic parsing, an approach that aims at both robustness and efficiency increase. Approximation techniques, to be applied at the level of language description, parsing strategy, and syntactic representation, have the same objective. Approximation at the level of syntactic representation is also known as underspecification, a traditional technique to deal with syntactic ambiguity. In this book new parsing technologies are collected that aim at attacking the problems of robustness and efficiency by exactly these techniques: the design of probabilistic grammars and efficient probabilistic parsing algorithms, approximation techniques applied to grammars and parsers to increase parsing efficiency, and techniques for underspecification and the integration of semantic information in the syntactic analysis to deal with massive ambiguity. The book gives a state-of-the-art overview of current research and development in parsing technologies. In its chapters we see how probabilistic methods have entered the toolbox of computational linguistics in order to be applied in both parsing theory and parsing practice. The book is both a unique reference for researchers and an introduction to the field for interested graduate students.

ANTONIS BOTINIS 1. 1 Background This introduction provides essential information about the structure and the objects of study of this volume. Following the introduction, fourteen papers which represent current research on intonation are organised into five thematic sections: (I) Overview of Intonation, (II) Prominence and Focus, (III) Boundaries and Discourse, (IV) Intonation Modelling, and (V) Intonation Technology. Within the sections the papers are arranged thematically, although several papers which deal with various aspects of intonation and prosody are basically intersectional. As the title indicates, "Intonation: Analysis, Modelling and Technology" is a contribution to the study of prosody, with major emphasis on intonation. Intonation and tonal themes are thus the central object of the volume, although temporal and dynamic aspects are also taken into consideration by a good number of papers. Although tonal and prosodic distinctions have been dealt with throughout man's literate history with reference to the study of language, for example by classical philosophers such as Plato and Aristotle, it is in recent decades that we have witnessed the most fertile growth in intonation studies, as with experimental phonetics and speech technology in general. As Rossi (this volume) points out, intonation research really began to blossom in the sixties with a multi­ fold increase in prosodic studies, reflected in contributions to the International Congress of Phonetic Sciences (ICPhS), and in the international literature.

In this book we address robustness issues at the speech recognition and natural language parsing levels, with a focus on feature extraction and noise robust recognition, adaptive systems, language modeling, parsing, and natural language understanding. This book attempts to give a clear overview of the main technologies used in language and speech processing, along with an extensive bibliography to enable topics of interest to be pursued further. It also brings together speech and language technologies often considered separately. Robustness in Language and Speech Technology serves as a valuable reference and although not intended as a formal university textbook, contains some material that can be used for a course at the graduate or undergraduate level.

This book is an introduction to the statistical analysis of word frequency distributions, intended for linguists, psycholinguistics, and researchers work­ ing in the field of quantitative stylistics and anyone interested in quantitative aspects of lexical structure. Word frequency distributions are characterized by very large numbers of rare words. This property leads to strange statisti­ cal phenomena such as mean frequencies that systematically keep changing as the number of observations is increased, relative frequencies that even in large samples are not fully reliable estimators ofpopulationprobabilities, and model parameters that emerge as functions of the text size. Special statistical techniques for the analysis of distributions with large numbers of rare events can be found in various technical journals. The aim of this book is to make these techniques more accessible for non-specialists. Chapter 1 introduces some basic concepts and notation. Chapter 2 describes non-parametricmethods for the analysis ofword frequency distributions. The next chapterdescribes in detail three parametricmodels, the lognormal model, the Yule-Simon Zipfian model, and the generalized inverse Gauss-Poisson model. Chapter 4 introduces the concept of mixture distributions. Chapter 5 explores the effectofnon-randomness inword use on the accuracy of the non­ parametric and parametric models, all of which are based on the assumption that words occur independently and randomly in texts. Chapter 6 presents examples of applications.

An Introduction to Text-to-Speech Synthesis is a comprehensive introduction to the subject. The author treats two areas of speech synthesis: Part I of the book concerns natural language processing and the inherent problems it presents for speech synthesis; Part II focuses on digital signal processing, with an emphasis on the concatenative approach. Both parts of the text guide the reader through the material in a step-by-step easy-to-follow way. This is the first book to treat the topic of speech synthesis from the perspective of two different engineering approaches. The book will be of interest to researchers and students in phonetics and speech communication, in both academia and industry.

In Marcus (1980), deterministic parsers were introduced. These are parsers which satisfy the conditions of Marcus's determinism hypothesis, i.e., they are strongly deterministic in the sense that they do not simulate non­ determinism in any way. In later work (Marcus et al. 1983) these parsers were modified to construct descriptions of trees rather than the trees them­ selves. The resulting D-theory parsers, by working with these descriptions, are capable of capturing a certain amount of ambiguity in the structures they build. In this context, it is not clear what it means for a parser to meet the conditions of the determinism hypothesis. The object of this work is to clarify this and other issues pertaining to D-theory parsers and to provide a framework within which these issues can be examined formally. Thus we have a very narrow scope. We make no ar­ guments about the linguistic issues D-theory parsers are meant to address, their relation to other parsing formalisms or the notion of determinism in general. Rather we focus on issues internal to D-theory parsers themselves.

This book is based on contributions to the Seventh European Summer School on Language and Speech Communication that was held at KTH in Stockholm, Sweden, in July of 1999 under the auspices of the European Language and Speech Network (ELSNET). The topic of the summer school was "Multimodality in Language and Speech Systems" (MiLaSS). The issue of multimodality in interpersonal, face-to-face communication has been an important research topic for a number of years. With the increasing sophistication of computer-based interactive systems using language and speech, the topic of multimodal interaction has received renewed interest both in terms of human-human interaction and human-machine interaction. Nine lecturers contri­ buted to the summer school with courses on specialized topics ranging from the technology and science of creating talking faces to human-human communication, which is mediated by computer for the handicapped. Eight of the nine lecturers are represented in this book. The summer school attracted more than 60 participants from Europe, Asia and North America representing not only graduate students but also senior researchers from both academia and industry.

This book is an introduction to the statistical analysis of word frequency distributions, intended for linguists, psycholinguistics, and researchers work­ ing in the field of quantitative stylistics and anyone interested in quantitative aspects of lexical structure. Word frequency distributions are characterized by very large numbers of rare words. This property leads to strange statisti­ cal phenomena such as mean frequencies that systematically keep changing as the number of observations is increased, relative frequencies that even in large samples are not fully reliable estimators ofpopulationprobabilities, and model parameters that emerge as functions of the text size. Special statistical techniques for the analysis of distributions with large numbers of rare events can be found in various technical journals. The aim of this book is to make these techniques more accessible for non-specialists. Chapter 1 introduces some basic concepts and notation. Chapter 2 describes non-parametricmethods for the analysis ofword frequency distributions. The next chapterdescribes in detail three parametricmodels, the lognormal model, the Yule-Simon Zipfian model, and the generalized inverse Gauss-Poisson model. Chapter 4 introduces the concept of mixture distributions. Chapter 5 explores the effectofnon-randomness inword use on the accuracy of the non­ parametric and parametric models, all of which are based on the assumption that words occur independently and randomly in texts. Chapter 6 presents examples of applications.

Linguists and engineers in Natural Language Processing tend to use electronic corpora more and more. Most research has long been limited to raw (unannotated) texts or to tagged texts (annotated with parts of speech only), but these approaches suffer from a word by word perspective. A new line of research involves corpora with richer annotations such as clauses and major constituents, grammatical functions and dependency links. The first parsed corpora were the English Lancaster treebank and Penn Treebank. New ones have recently been developed for other languages. This book: provides a state of the art on work being done with parsed corpora; gathers 21 papers on building and using parsed corpora raising many relevant questions; deals with a variety of languages and a variety of corpora; is for those working in linguistics, computational linguistics, natural language, syntax, and grammar.

Linguists and engineers in Natural Language Processing tend to use electronic corpora more and more. Most research has long been limited to raw (unannotated) texts or to tagged texts (annotated with parts of speech only), but these approaches suffer from a word by word perspective. A new line of research involves corpora with richer annotations such as clauses and major constituents, grammatical functions and dependency links. The first parsed corpora were the English Lancaster treebank and Penn Treebank. New ones have recently been developed for other languages. This book: provides a state of the art on work being done with parsed corpora; gathers 21 papers on building and using parsed corpora raising many relevant questions; deals with a variety of languages and a variety of corpora; is for those working in linguistics, computational linguistics, natural language, syntax, and grammar.

This work is of relevance to researchers and programme developers in the field of Machine Translation and especially Example-Based Machine Translation, bilingual text processing and cross-linguistic information retrieval. It is also of interest to translation technologists and localisation professionals. Recent Advances in Example-Based Machine Translation fills a void, because it is the first book to tackle the issue of EBMT in depth. It gives a state-of-the-art overview of EBMT techniques and provides a coherent structure in which all aspects of EBMT are embedded. Its contributions are written by long-standing researchers in the field of MT in general, and EBMT in particular. This book can be used in graduate-level courses in machine translation and statistical NLP.

Recent Advances in Example-Based Machine Translation is of relevance to researchers and program developers in the field of Machine Translation and especially Example-Based Machine Translation, bilingual text processing and cross-linguistic information retrieval. It is also of interest to translation technologists and localisation professionals.Recent Advances in Example-Based Machine Translation fills a void, because it is the first book to tackle the issue of EBMT in depth. It gives a state-of-the-art overview of EBMT techniques and provides a coherent structure in which all aspects of EBMT are embedded. Its contributions are written by long-standing researchers in the field of MT in general, and EBMT in particular. This book can be used in graduate-level courses in machine translation and statistical NLP.

The origins of this book arise from the highly successful second SIGdial Workshop on Discourse and Dialogue that was held in September 2001 in con­ junction with Eurospeech 2001. The original workshop proceedings consisted of 29 papers selected from 57 submissions, an exceptionally high number of submissions for a two day workshop. This book includes extended versions of 12 papers originally presented at the workshop. In addition, 4 other invited papers on major themes in discourse and dialogue research are included. There are three main themes addressed by the papers in this collection: (1) corpus annotation and analysis; (2) method­ ologies for construction of dialogue systems; and (3) perspectives on various key theoretical issues including communicative intention, context-based gen­ eration, and modeling of discourse structure. However, because of the very nature of discourse and dialogue research that often requires researchers to tackle several issues in one piece of work, we have chosen to order the papers alphabetically by author rather than try to create artificial thematic sections. We believe this collection provides a concise yet reasonably comprehensive snapshot of major research themes in discourse and dialogue. We hope that readers will benefit greatly from this collection.