We propose and compare a number of metrics to capture the degree to which words are restricted in the contexts in which they can occur. We re-frame the problem of contextual restrictedness, and introduce the use of vector space models based on syntactic dependencies. We show that our most successful metric, residualized entropy, is quite successful in selecting highly collocationally restricted...

In this work, we describe a methodology based on the Stochastic Finite State Transducers paradigm for Spoken Language Understanding (SLU) for obtaining concept graphs from word graphs. In the edges of these concept graphs, both semantic and lexical information are represented. This makes these graphs a very useful representation of the information for SLU. The best path in these concept graphs ...

Information extraction in the sense of the Message Understanding Conferences has been traditionally defined as the extraction of information from a text in the form of text strings and processed text strings which are placed into slots labeled to indicate the kind of information that can fill them. So, for example, a slot labeled NAME would contain a name string taken directly out of the text o...

This paper investigates two strategies for improving coreference resolution: (1) training separate models that specialize in particular types of mentions (e.g., pronouns versus proper nouns) and (2) using a ranking loss function rather than a classification function. In addition to being conceptually simple, these modifications of the standard single-model, classification-based approach also de...

This paper reflects about the kinds of morphological, syntactic, semantic, and pragmatic knowledge needed to process ill-formed input. We conclude that an excellent start on processing ill—formed input has been exemplified in a number of concrete implementations, but that a substantial amount of fundamental work must still be done if our systems are to understand language robustly to the degree...

the purported mechanism underlying the development and progression of cervicogenic headache (ceh) is the convergence of sensory inputs at the trigeminocervical nucleus. this mechanism explains the radiation of pain from the neck or the occipitonuchal area and its spread to the oculo-fronto-temporal region; it also explains the recurrent headaches caused by improper neck postures or external pre...

Implicative verbs (e.g. manage) entail their complement clauses, while non-implicative verbs (e.g. want) do not. For example, while managing to solve the problem entails solving the problem, no such inference follows from wanting to solve the problem. Differentiating between implicative and non-implicative verbs is therefore an essential component of natural language understanding, relevant to ...

The beneets of the eeective creation of Information Extraction (IE) in the last ten years, driven by the ARPA TIPSTER programme and the associated MUC evaluations, have been enormous, but it must now be time to ask what research issues face the systems we have built and what we should do next. We suggest that there are two classes of important research issues: those requiring detailed comparati...