Anaphora

"This was a different man," said Mr Welbecker. "Listen! This man was called Hamlet and his uncle had killed his father because he wanted to marry his mother.""What did he want to marry his mother for?" said William. "I've never heard of anyone wanting to marry their mother."*In almost any conversation the meaning of what is said depends on the listener seeing how some words refer to what has already been said, and that others must be related to the characteristics of time, place, or person of the situation around which the conversation revolves. These modes ofreference, anaphora and deixis respectively, involve surprisingly complicated cognitive and syntactic processes, which people (normally) perform easily and unerringly. But they present formidable problems for the linguist and cognitive scientist trying to explain precisely how comprehension isachieved. Anaphora and deixis are thus a central research focus in syntactic theory, while understanding and modelling their operation in discourse are important targets in computational linguistics and cognitive science. In this ambitious work, Francis Cornish sets out an original theory ofanaphora and deixis, and proposes a new and elegant theoretical model to represent the transfer of meaning in discourse.Dr Cornish considers anaphoric reference in discourse from both psychological and linguistic perspectives. He argues that anaphora and deixis are essentially parts of integrative discourse procedures that facilitate the linking of representations held in working memory. He brings together work bylinguists, formal semanticists, psychologists, and researchers in artificial intelligence, as well as drawing on his own extensive experimental work on a variety of corpora of different genres in French and English.Anaphora, Discourse, and Understanding will interest researchers and advanced students in a variety of fields within and outside linguistics, including cognitive science, artificial intelligence, syntactic theory, formal semantics, and the analysis of discourse.[* from William - The Pirate by Richmal Crompton, London, Macmillan, 1932]

I n order to appreciate properly what we are doing in this book it is necessary to realize that our approach to linguistic theorizing differs from the prevailing views. Our approach can be described by indicating what distinguishes it from the methodological ideas current in theoretical linguistics, which I consider seriously misguided. Linguists typically construe their task in these days as that of making exceptionless generalizations from particular examples. This explanatory strategy is wrong in several different ways. It presupposes that we can have "intuitions" about particular examples, usually examples invented by the linguist himself or herself, reliable and sharp enough to serve as a basis of sharp generalizations. It also presupposes that we cannot have equally reliable direct access to general linguistic regularities. Both assumptions appear to me extremely dubious, and the first of them has in effect been challenged by linguists like Dwight Bol inger. There is also some evidence that the degree of unanimity among linguists is fairly low when it comes to less clear cases, even in connection with such relatively simple questions as grammaticality (acceptability). For this reason we have tried to rely more on quotations from contemporary fiction, newspapers and magazines than on linguists' and philosophers' ad hoc examples. I also find it strange that some of the same linguists as believe that we all possess innate ideas about general characteristics of humanly possible grammars assume that we can have access to them only via their particular consequences.

Anaphora processing is a central topic in the study of natural language and has long been the object of research in a wide range of disciplines. The correct interpretation of anaphora has also become increasingly important for real-world natural language processing applications, including machine translation, automatic abstracting, information extraction and question answering. This volume provides a unique overview of the processing of anaphora from a multi- and inter-disciplinary angle. It will be of interest and practical use to readers from fields as diverse as theoretical linguistics, corpus linguistics, computational linguistics, computer science, natural language processing, artificial intelligence, human language technology, psycholinguistics, cognitive science and translation studies. The readership includes but is not limited to university lecturers, researchers, postgraduate and senior undergraduate students.

Teaching computers to solve language problems is one of the major challenges of natural language processing. There is a large amount of interesting research devoted to this field. This book fills an existing gap in the literature with an up-to-date survey of the field, including the author’s own contributions. A number of different fields overlap in anaphora resolution – computational linguistics, natural language processing (NLP), grammar, semantics, pragmatics, discourse analysis and artificial intelligence. This book begins by introducing basic notions and terminology, moving onto early research methods and approaches, recent developments and applications, and future directions. It addresses various issues related to the practical implementation of anaphora systems, such as rules employed, algorithms implemented or evaluation techniques used. This is an ideal reference book for students and researchers in this particular area of computational linguistics. Since anaphora resolution is vital for the development of any practical NLP system, the book will be of interest to readers from both academia and industry.

Intuitively, it is clear why languages have anaphoric relations: anaphora reduces redundancy, thereby shortening (and hence simplifying) sentences. In order for this simplification to be possible, however, it is necessary that the speaker of a language be able to identify correctly the elements participating in an anaphoric relation and to determine correctly the meaning of the anaphor on the basis of meaning of the antecedent. If a grammar is to reflect the linguistic competence of a native speaker of a language, it must include mechanisms of associating anaphor and antecedent. In this volume the following questions will be considered: What sorts of mechanisms are best suited for representing anaphora in a grammar? What are the conditions on the rule(s) associating anaphors with antecedents? Do the various cases of anaphora form a linguistically significant class of phenomena, and, if so, how can the grammar capture this fact? And what do these answers entail for linguistic theory?

This book lays out a path leading from the linguistic and cognitive basics, to classical rule-based and machine learning algorithms, to today’s state-of-the-art approaches, which use advanced empirically grounded techniques, automatic knowledge acquisition, and refined linguistic modeling to make a real difference in real-world applications. Anaphora and coreference resolution both refer to the process of linking textual phrases (and, consequently, the information attached to them) within as well as across sentence boundaries, and to the same discourse referent. The book offers an overview of recent research advances, focusing on practical, operational approaches and their applications. In part I (Background), it provides a general introduction, which succinctly summarizes the linguistic, cognitive, and computational foundations of anaphora processing and the key classical rule- and machine-learning-based anaphora resolution algorithms. Acknowledging the central importance of shared resources, part II (Resources) covers annotated corpora, formal evaluation, preprocessing technology, and off-the-shelf anaphora resolution systems. Part III (Algorithms) provides a thorough description of state-of-the-art anaphora resolution algorithms, covering enhanced machine learning methods as well as techniques for accomplishing important subtasks such as mention detection and acquisition of relevant knowledge. Part IV (Applications) deals with a selection of important anaphora and coreference resolution applications, discussing particular scenarios in diverse domains and distilling a best-practice model for systematically approaching new application cases. In the concluding part V (Outlook), based on a survey conducted among the contributing authors, the prospects of the research field of anaphora processing are discussed, and promising new areas of interdisciplinary cooperation and emerging application scenarios are identified. Given the book’s design, it can be used both as an accompanying text for advanced lectures in computational linguistics, natural language engineering, and computer science, and as a reference work for research and independent study. It addresses an audience that includes academic researchers, university lecturers, postgraduate students, advanced undergraduate students, industrial researchers, and software engineers.

This book covers anaphora resolution for the English language from a linguistic and computational point of view. First, a definition of anaphors that applies to linguistics as well as information technology is given. On this foundation, all types of anaphors and their characteristics for English are outlined. To examine how frequent each type of anaphor is, a corpus of different hypertexts has been established and analysed with regard to anaphors. The most frequent type are non-finite clause anaphors - a type which has never been investigated so far. Therefore, the potential of non-finite clause anaphors are further explored with respect to anaphora resolution. After presenting the fundamentals of computational anaphora resolution and its application in text retrieval, rules for resolving non-finite clause anaphors are established. Therefore, this book shows that a truly interdisciplinary approach can achieve results which would not have been possible otherwise.

Discourse anaphora is a challenging linguistic phenomenon that has given rise to research in fields as diverse as linguistics, computational linguistics and cognitive science. Because of the diversity of approaches these fields bring to the anaphora problem, the editors of this volume argue that there needs to be a synthesis, or at least a principled attempt to draw the differing strands of anaphora research together. The selected papers in this volume all contribute to the aim of synthesis and were selected to represent the growing importance of corpus-based and computational approaches to anaphora description, and to developing natural language systems for resolving anaphora in natural language.