Marc B. Vilain

A model-theoretic coreference scoring scheme

This note describes a scoring scheme for the coreference task in MUC6. It improves on the original approach by: (1) grounding the scoring scheme in terms of a model; (2) producing more intuitive recall and precision scores; and (3) not requiring explicit computation of the transitive closure of coreference. The principal conceptual difference is that we have moved from a syntactic scoring model based on following coreference links to an approach defined by the model theory of those links.

Constraint propagation algorithms for temporal reasoning: a revised report

This paper revises and expands upon a paper presented by two of the present authors at AAAI 1986 [Vilain & Kautz 1986]. As with the original, this revised document considers computational aspects of interval-based and point-based temporal representations. Computing the consequences of temporal assertions is shown to be computationally intractable in the interval-based representation, but not in the point-based one. However, a fragment of the interval language can be expressed using the point language and benefits from the tractability of the latter. The present paper departs from the original primarily in correcting claims made there about the point algebra, and in presenting some closely related results of van Beek [1989].

MITRE: description of the Alembic system used for MUC-6

As with several other veteran MUC participants, MITREs Alembic system has undergone a major transformation in the past two years. The genesis of this transformation occurred during a dinner conversation at the last MUC conference, MUC-5. At that time, several of us reluctantly admitted that our major impediment towards improved performance was reliance on then-standard linguistic models of syntax. We knew we would need an alternative to traditional linguistic grammars, even to the somewhat non-traditional categorial pseudo-parser we had in place at the time. The problem was, which alternative?

Finite-state phrase parsing by rule sequences

We present a novel approach to parsing phrase grammars based on Eric Brills notion of rule sequences. The basic framework we describe has somewhat less power than a finite-state machine, and yet achieves high accuracy on standard phrase parsing tasks. The rule language is simple, which makes it easy to write rules. Further, this simplicity enables the automatic acquisition of phrase-parsing rules through an error-reduction strategy.

Some Properties of Preposition and Subordinate Conjunction Attachments

Determining the attachments of prepositions and subordinate conjunctions is a key problem in parsing natural language. This paper presents a trainable approach to making these attachments through transformation sequences and error-driven learning. Our approach is broad coverage, and accounts for roughly three times the attachment cases that have previously been handled by corpus-based techniques. In addition, our approach is based on a simplified model of syntax that is more consistent with the practice in current state-of-the-art language processing systems. This paper sketches syntactic and algorithmic details, and presents experimental results on data sets derived from the Penn Treebank. We obtain an attachment accuracy of 75.4% for the general case, the first such corpus-based result to be reported. For the restricted cases previously studied with corpusbased methods, our approach yields an accuracy comparable to current work (83.1%).

Phrase parsing with rule sequence processors: an application to the shared CoNLL task

For several years, chunking has been an integral part of MITREs approach to information extraction. Our work exploits chunking in two principal ways. First, as part of our extraction system (Alembic) (Aberdeen et al., 1995), the chunker delineates descriptor phrases for entity extraction. Second, as part of our ongoing research in parsing, chunks provide the first level of a stratified approach to syntax - the second level is defined by grammatical relations, much as in the SPARKLE effort (Carroll et al., 1997).

Inferential Information Extraction

This paper is concerned with an outlook on information extraction (IE) that is steeped to a large extent in the traditional semantic notion of inferential reasoning. We make the case for a continued presence of inferential methods in IE, despite the ongoing trend towards simpler extraction processing models. We demonstrate the role of this kind of inference in the Alembic message understanding system, and also discuss the upstream syntactic processing that enables this. We present the finite-state parsing models that until recently have served this role, and cover at some length the issues of semantic interpretation that they require. As a pointer towards work to come, we also mention briefly our work in progress on parsing via grammatical relations, an approach that we hope will add great generality to our extraction framework.

MITRE: DESCRIPTION OF THE ALEMBIC SYSTEM AS USED IN MET

Alembic is a comprehensive information extraction system that has been applied to a range of tasks. These include the now-standard components of the formal MUC evaluations: name tagging (NE in MUC-6), name normalization (TE), and template generation (ST). The system has also been exploited to help segment and index broadcast video and was used for early experiments on variants of the co-reference identification task. (For details, see [1].)

The relation-based knowledge representation of King Kong

This paper presents an overview of the knowledge representation facilities of King Kong a transportable natural language system. The thrust of the paper is towards demonstrating how the particulars of Kongs representation language support the processing of key phenomena of natural language. To this extent, we cover Kongs terminological hierarchies; the logical language in which Kong encodes utterance meanings; and the query evaluator that interfaces Kong to expert system back ends. We focus particularly on King Kongs innovative treatment of relations, as this treatment provides the system with much of its language analysis strengths.

Entity Extraction is a Boring Solved Problem---Or is it?

This paper presents empirical results that contradict the prevailing opinion that entity extraction is a boring solved problem. In particular, we consider data sets that resemble familiar MUC/ACE data, and report surprisingly poor performance for both commercial and research systems. We then give an error analysis that suggests research challenges for entity extraction that are neither boring nor solved.