Tracy Holloway King

Parsing the Wall Street Journal using a Lexical-Functional Grammar and Discriminative Estimation Techniques

We present a stochastic parsing system consisting of a Lexical-Functional Grammar (LFG), a constraint-based parser and a stochastic disambiguation model. We report on the results of applying this system to parsing the UPenn Wall Street Journal (WSJ) treebank. The model combines full and partial parsing techniques to reach full grammar coverage on unseen data. The treebank annotations are used to provide partially labeled data for discriminative statistical estimation using exponential models. Disambiguation performance is evaluated by measuring matches of predicate-argument relations on two distinct test sets. On a gold standard of manually annotated f-structures for a subset of the WSJ treebank, this evaluation reaches 79% F-score. An evaluation on a gold standard of dependency relations for Brown corpus data achieves 76% F-score.

The Parallel Grammar project

We report on the Parallel Grammar (ParGram) project which uses the XLE parser and grammar development platform for six languages: English, French, German, Japanese, Norwegian, and Urdu.

Activity-based serendipitous recommendations with the Magitti mobile leisure guide

This paper presents a context-aware mobile recommender system, codenamed Magitti. Magitti is unique in that it infers user activity from context and patterns of user behavior and, without its user having to issue a query, automatically generates recommendations for content matching. Extensive field studies of leisure time practices in an urban setting (Tokyo) motivated the idea, shaped the details of its design and provided data describing typical behavior patterns. The paper describes the fieldwork, user interface, system components and functionality, and an evaluation of the Magitti prototype.

Statistical sentence condensation using ambiguity packing and stochastic disambiguation methods for Lexical-Functional Grammar

We present an application of ambiguity packing and stochastic disambiguation techniques for Lexical-Functional Grammars (LFG) to the domain of sentence condensation. Our system incorporates a linguistic parser/generator for LFG, a transfer component for parse reduction operating on packed parse forests, and a maximum-entropy model for stochastic output selection. Furthermore, we propose the use of standard parser evaluation methods for automatically evaluating the summarization quality of sentence condensation systems. An experimental evaluation of summarization quality shows a close correlation between the automatic parse-based evaluation and a manual evaluation of generated strings. Overall summarization quality of the proposed system is state-of-the-art, with guaranteed grammaticality of the system output due to the use of a constraint-based parser/generator.

Determiner agreement and noun conjunction

Determiner-noun agreement in English and many other languages appears to be straightforwardly describable; singular determiners go with singular nouns, and plural determiners go with plural nouns. The situation is more complicated with coordinated nouns, however, since unexpected agreement patterns often result. Our theory makes the correct predictions for English and other languages by combining two crucial insights: the dual nature of agreement features inside the noun phrase (Kathol 1999; Sadler 1999, 2003; Wechsler & Zlatic 2000, 2003) and the distinction between distributive and nondistributive features in coordination (Dalrymple & Kaplan 2000).

Deverbal Nouns in Knowledge Representation

Deverbal nouns pose serious challenges for knowledge-representation systems. We present a method of canonicalizing deverbal noun representations, relying on a rich lexicon of verb subcategorization frames, the WordNet database, a large finite-state network for derivational morphology and a series of heuristics for mapping deverbal arguments onto the arguments of corresponding verbs.1

Urdu in a parallel grammar development environment

In this paper, we report on the role of the Urdu grammar in the Parallel Grammar (ParGram) project (Butt, M., King, T. H., Niño, M.-E., & Segond, F. (1999). A grammar writer’s cookbook. CSLI Publications; Butt, M., Dyvik, H., King, T. H., Masuichi, H., & Rohrer, C. (2002). ‘The parallel grammar project’. In: Proceedings of COLING 2002, Workshop on grammar engineering and evaluation, pp. 1–7). The Urdu grammar was able to take advantage of standards in analyses set by the original grammars in order to speed development. However, novel constructions, such as correlatives and extensive complex predicates, resulted in expansions of the analysis feature space as well as extensions to the underlying parsing platform. These improvements are now available to all the project grammars.

Adapting existing grammars: the XLE experience

We report on the XLE parser and grammar development platform (Maxwell and Kaplan, 1993) and describe how a basic Lexical Functional Grammar for English has been adapted to two different corpora (newspaper text and copier repair tips).

The Rules of Redaction: Identify, Protect, Review (and Repeat)

Frequent data leak reports in the press attest to the difficulty of identifying and protecting sensitive content. Redaction is particularly challenging because it seeks to protect a document by selectively removing only the sensitive parts. The prototype system described here addresses some current technology limitations.

Indeterminacy by underspecification

We examine the formal encoding of feature indeterminacy, focussing on case indeterminacy as an exemplar of the phenomenon. Forms that are indeterminately specified for the value of a feature can simultaneously satisfy conflicting requirements on that feature and thus are a challenge to constraint-based formalisms which model the compatibility of information carried by linguistic items by combining or integrating that information. Much previous work in constraint-based formalisms has sought to provide an analysis of feature indeterminacy by departing in some way from vanilla assumptions either about feature representations or about how compatibility is checked by integrating information from various sources. In the present contribution we argue instead that a solution to the range of issues posed by feature indeterminacy can be provided in a vanilla feature-based approach which is formally simple, does not postulate special structures or objects in the representation of case or other indeterminate features, and requires no special provision for the analysis of coordination. We view the value of an indeterminate feature such as case as a complex and possibly underspecified feature structure. Our approach correctly allows for incremental and monotonic refinement of case requirements in particular contexts. It uses only atomic boolean-valued features and requires no special mechanisms or additional assumptions in the treatment of coordination or other phenomena to handle indeterminacy. Our account covers the behaviour of both indeterminate arguments and indeterminate predicates, that is, predicates placing indeterminate requirements on their arguments.