Anne N. De Roeck

A morphologically sensitive clustering algorithm for identifying Arabic roots

We present a clustering algorithm for Arabic words sharing the same root. Root based clusters can substitute dictionaries in indexing for IR. Modifying Adamson and Boreham (1974), our Two-stage algorithm applies light stemming before calculating word pair similarity coefficients using techniques sensitive to Arabic morphology. Tests show a successful treatment of infixes and accurate clustering to up to 94.06% for unedited Arabic text samples, without the use of dictionaries.

Extracting Semistructured Data - Lessons Learnt

The Yellow Pages Assistant (YPA) is a natural language dialogue system which guides a user through a dialogue in order to retrieve addresses from the Yellow Pages. Part of the work in this project is concerned with the construction of a Backend, i.e. the database extracted from the raw input text that is needed for the online access of the addresses. Here we discuss some aspects involved in this task as well as report on experiences which might be interesting for other projects as well.

Helpful answers to modal and hypothetical questions

This paper describes a computational pragmatic model which is geared towards providing helpful answers to modal and hypothetical questions. The work brings together elements from formal semantic theories on modality and question answering, defines a wider, pragmatically flavoured, notion of answerhood based on non-monotonic inference and develops a notion of context, within which aspects of more cognitively oriented theories, such as Relevance Theory, can be accommodated. The model has been implemented. The research was funded By ESRC grant number R000231279.

The YPA - An Assistant for Classified Directory Enquiries

The Ypa is a directory enquiry system which allows a user to access advertiser information in classified directories [1]. It converts semi-structured data in the Yellow Pages machine readable classified directories into a set of indices appropriate to the domain and task, and converts natural language queries into filled slot and filler structures appropriate for queries in the domain. The generation of answers requires a domain-dependent query construction step, connecting the indices and the slot and fillers. The Ypa illustrates an unusual but useful intermediate point between information retrieval and logical knowledge representation.

A Belief-Centered Treatment of Pragmatic Presupposition

Traditional, semantic, accounts of presupposition centre on truth conditional treatments relying on entailments. In contrast, our work concentrates on the investigation of presupposition as a pragmatic phenomenon which interacts with agents’ beliefs. Several accounts (e.g., [1],[2],[5],[6],[7],[8],[23]) have taken contextual factors into account. However, they run into problems caused by the consequences of the basic possible world based notions of propositionhood they deploy, which prevent the formulation of partial, revisable information states.

Generating English paraphrases from relational query expressions

Abstract Of the many problems facing the casual user of a data–base enquiry system probably the most difficult is gaining a competent understanding of the associated query language. Given that he manages to construct a well–formed query expression there is no guarantee that it exactly reflects the original question. The research described here concerns the design of an interpreter from a formal query language to natural language to aid query verification in a relational data–base environment. The system is being developed to work in conjunction with the ICL Natural Language enquiry interface NEL which translates English query expressions into the formal query language QUERYMASTER. The requirements of a natural–language paraphraser are first discussed and the nature of an intermediate representation is defined and motivated with respect to an applied relational calculus. Consideration is then given to choosing a suitable underlying framework with which to underpin the practical work and the choice of Lexic...

Natural language front-ends to databases: design and the customisation bottleneck

NLFE to databases have failed in a commercial context, largely because of two reasons. Current approaches to the management of ambiguity by relying on inference over a world model create ungoing customisation requirements. Furthermore the design of NLFEs is subject to constraints which research in CL/ NLP does not address. In particular, standard parsing techniques (including robust ones) require complete lexica and cannot be deployed because new data would create a constant need for dictionary update. The SQUIRREL [1] system (SERC Grant GR/E/ 69485) addresses some of these problems: its design reduces customisation effort as words are interpreted without reference to world models. The lexicon is assumed to be incomplete: unknown words are given interpretations by exploiting typing information contained in the datamodel. In addition, SQUIRREL demonstrates that NLFEs can allow for interrogation of integrity constraints, usually invisible to users. It is important to note that no new aspects of standard database management systems are involved SQUIRREl. intends to explore to what extent the state of the art in NLP/CL and Formal Semantics can be exploited in the design of NLFE to relational databases, under constraints imposed by good sofware engineering protocol. It aims to develop a modular, portable design, to plug in to public domain database technology, requiring minimal customisation. SQUIRREl. consists of a series of mappings translating NL expressions into SQL. Its highly modular design allows parts of the system to be ported without affecting other parts. Expressions in English are assigned syntactic and semantic representations on the basis of a lexicon and a context-free feature ba~d grammar. The lexicon is incomplete: unknown words are assigned tentative categories by the (bi-directional chart) parser. Syntactic and semantic rules operate in tandem. Semantic representations are cast in Property Theory (PD [2], delivering intensional objects. These are assigned extensions in the form of first order logic (FOL) expressions. So far, the representations are independent from the domain model of any database in question. The FOL expressions are translated into the domain relational calculus (DRC), by rules exploiting the logical structure of the FOL formulae, and a domain model. The resulting expressions are translated into SQL by a simple syntactic transduction. The design offers several cut-off points at which modules can be re-deployed. The lexicon and granunar, currently written for a subset of English, can readily be customised for any language for which a context-free feature based grammar exists. The step via PT offers a second point where the system can be deployed to applications other than database interfaces. The mapping into the DRC makes it possible to port the system to any relational query language. The real advance made in this system is the economy of its datamodel. It sets out how each word in the dictionary is to be understood w.r.t, the current database by direct mapping: no world knowledge or inference is required. Unknown words are filled in by typing constraints associated with domains in the datamodel. No loss of expressiveness is entailed: this is hardly surprising as all a world model would seek to do is to (i) exaggerate ambiguity w.r.t, how a user might perceive the world, in order to (ii) reduce that ambiguity w.r.t. what the current database can provide. Under this view, step (i) is totally superfluous. The resulting gain in customisation effort is paramount. SQUIRREI.s ambiguity management strategy is to offer users a choice between all interpretations that have survived the mapping into SQL. Note that at each stage in the mapping, alternative representations may emerge, or existing ones may die off. The most powerful disambiguation tool is the exploitation Of typing constraints associated with the database itself.