Patrick Hanks

Word Association Norms, Mutual Information, and Lexicography

The term word association is used in a very particular sense in the psycholinguistic literature. (Generally speaking, subjects respond quicker than normal to the word nurse if it follows a highly associated word such as doctor. ) We will extend the term to provide the basis for a statistical description of a variety of interesting linguistic phenomena, ranging from semantic relations of the doctor/nurse type (content word/content word) to lexico-syntactic co-occurrence constraints between verbs and prepositions (content word/function word). This paper will propose an objective measure based on the information theoretic notion of mutual information, for estimating word association norms from computer readable corpora. (The standard method of obtaining word association norms, testing a few thousand subjects on a few hundred words, is both costly and unreliable.) The proposed measure, the association ratio, estimates word association norms directly from computer readable corpora, making it possible to estimate norms for tens of thousands of words.

Lexical analysis : norms and exploitations

In Lexical Analysis, Patrick Hanks offers a wide-ranging empirical investigation of word use and meaning in language. The book fills the need for a lexically based, corpus-driven theoretical approach that will help people understand how words go together in collocational patterns and constructions to make meanings. Such an approach is now possible, Hanks writes, because of the availability of new forms of evidence (corpora, the Internet) and the development of new methods of statistical analysis and inferencing. Hanks offers a new theory of language, the Theory of Norms and Exploitations (TNE), which makes a systematic distinction between normal and abnormal usage -- between rules for using words normally and rules for exploiting such norms in metaphor and other creative use of language. Using hundreds of carefully chosen citations from corpora and other texts, he shows how matching each use of a word against established contextual patterns plays a large part in determining the meaning of an utterance. His goal is to develop a coherent and practical lexically driven theory of language that takes into account the immense variability of everyday usage and that shows that this variability is rule governed rather than random. Such a theory will complement other theoretical approaches to language, including cognitive linguistics, construction grammar, generative lexicon theory, priming theory, and pattern grammar.

Do Word Meanings Exist

1. IntroductionMy contribution to this discussion is to attempt to spread a little radical doubt.Since I have spent over 30 years of my life writing and editing monolingualdictionary definitions, it may seem rather odd that I should be asking, do wordmeanings exist? The question is genuine, though: prompted by some puzzling factsabout the data that is now available in the form of machine-readable corpora. Iam not the only lexicographer to be asking this question after studying corpusevidence. Sue Atkins, for example, has said “I don’t believe in word meanings”(personal communication).It is a question of fundamental importance to the enterprise of sensedisambiguation. If senses don’t exist, then there is not much point in tryingto ‘disambiguate’ them – or indeed do anything else with them. The veryterm disambiguate presupposes what Fillmore (1975) characterized as “checklisttheories of meaning.” Here I shall reaffirm the argument, on the basis of recentwork in corpus analysis, that checklist theories in their current form are at bestsuperficial and at worst misleading. If word meanings do exist, they do not exist asa checklist. The numbered lists of definitions found in dictionaries have helped tocreate a false picture of what really happens when language is used.Vagueness and redundancy – features which are not readily compatible with achecklist theory – are important design features of natural language, which mustbe taken into account when doing serious natural language processing. Wordsare so familiar to us, such an everyday feature of our existence, such an integraland prominent component of our psychological makeup, that it’s hard to see whatmysterious, complex, vague-yet-precise entities meanings are.2. Common SenseThe claim that word meaning is mysterious may seem counterintuitive. To take atime-worn example, it seems obvious that the nounbank has at least two senses:‘slope of land alongside a river’ and ‘financial institution’. But this line of argumentis a honeytrap. In the first place, these are not, in fact, two senses of a single word;

Automated induction of sense in context

In this paper, we introduce a model for sense assignment which relies on assigning senses to the contexts within which words appear, rather than to the words themselves. We argue that word senses as such are not directly encoded in the lexicon of the language. Rather, each word is associated with one or more stereotypical syntagmatic patterns, which we call selection contexts. Each selection context is associated with a meaning, which can be expressed in any of various formal or computational manifestations. We present a formalism for encoding contexts that help to determine the semantic contribution of a word in an utterance. Further, we develop a methodology through which such stereotypical contexts for words and phrases can be identified from very large corpora, and subsequently structured in a selection context dictionary, encoding both stereotypical syntactic and semantic information. We present some preliminary results.

A Diagnostic Database of American Personal Names

Abstract Computational analysis of pairings between surnames and forenames in “A Diagnostic Database of American Names” reveals a number of significant associations. These can be used as evidence to help pinpoint the most likely source language of an unfamiliar and unresearched surname, and to provide a framework for historical and genealogical research. The data can also be studied geographically since a number of surnames are associated with particular states or regions. A study of present-day distribution of surnames can be correlated’ with historical and other evidence regarding settlement patterns.

Common Sense About Word Meaning: Sense in Context

We present a new approach to determining the meaning of words in text, which relies on assigning senses to the contexts within which words occur, rather than to the words themselves. A preliminary version of this approach is presented in Pustejovsky, Hanks and Rumshisky (2004, COLING). We argue that words senses are not directly encoded in the lexicon of a language, but rather that each word is associated with one or more stereotypical syntagmatic patterns. Each pattern is associated with a meaning, which can be expressed in a formal way as a resource for any of a variety of computational applications.

Compiling a Monolingual Dictionary for Native Speakers

This article gives a survey of the main issues confronting the compilers of monolin- gual dictionaries in the age of the Internet. Among others, it discusses the relationship between a lexical database and a monolingual dictionary, the role of corpus evidence, historical principles in lexicography vs. synchronic principles, the instability of word meaning, the need for full vocabu- lary coverage, principles of definition writing, the role of dictionaries in society, and the need for dictionaries to give guidance on matters of disputed word usage. It concludes with some questions about the future of dictionary publishing.

Mechanisms of Meaning

Word meaning is at best a very vague phenomenon – some lexicographers, including the present writer, have gone so far as to claim that word meanings do not exist. So how is it possible that people can achieve precision in the meaning of their utterances? And how is it possible to use language creatively, to talk about new concepts or to talk about old concepts in new ways? The answer is surprising; it calls into question most previous work in computational linguistics on the so-called ‘word sense disambiguation problem’, which, I shall argue, is still unresolved because it is based on unsound theoretical assumptions. If word senses do not exist, they surely cannot be disambiguated (or processed in any other way). The hypothesis to be explored in this paper is that meanings are associated with the phraseological patterns associated with each word in normal usage, rather than with words themselves.

Adam Kilgarriff’s Legacy to Computational Linguistics and Beyond

The 2016 CICLing conference was dedicated to the memory of Adam Kilgarriff who died the year before. Adam leaves behind a tremendous scientific legacy and those working in computational linguistics, other fields of linguistics and lexicography are indebted to him. This paper is a summary review of some of Adam’s main scientific contributions. It is not and cannot be exhaustive. It is written by only a small selection of his large network of collaborators. Nevertheless we hope this will provide a useful summary for readers wanting to know more about the origins of work, events and software that are so widely relied upon by scientists today, and undoubtedly will continue to be so in the foreseeable future.

Corpus Patterns for Semantic Processing

This tutorial presents a corpus-driven, pattern-based empirical approach to meaning representation and computation. Patterns in text are everywhere, but techniques for identifying and processing them are still rudimentary. Patterns are not merely syntactic but syntagmatic: each pattern identifies a lexico-semantic clause structure consisting of a predicator (verb or predicative adjective) together with open-ended lexical sets of collocates in different clause roles (subject, object, prepositional argument, etc.). If NLP is to make progress in identifying and processing text meaning, pattern recognition and collocational analysis will play an essential role, because: