This paper describes the analysis component of a natural language dialogue system called FALIN. The most important features of the system are its modularity and the combination of an object-oriented semantics with a constraint-based grammar formalism. I also illustrate how constraints in the form of descriptor schemata can be used to capture general correspondences between syntactic and semantic structure and argue that the contribution of contextual factors to interpretation to some extent can be handled within the same framework.

This paper describes a system for automatic sentence synthesis (or “generation”) from semantic representations. The system is implemented in Interlisp as an extension of D-PATR, Lauri Karttunen's environment for the development of unification-based grammars (see Karttunen 1986, Carlson & Linden 1987).

The fact that a lexeme can appear in various forms causes problems in information retrieval. As a solution to this problem, we have developed methods for automatic root lemmatization, automatic truncation and automatic splitting of compound words. All the methods have as their basis a set of rules which contain information regarding inflected and derived forms of words – and not a dictionary. The methods have been tested on several collections of texts, and have produced very good results. By controlled experiments in text retrieval, we have studied the effects on search results. These results show that both the method of automatic root lemmatization and the method of automatic truncation make a considerable improvement on search quality. The experiments with splitting of compound words did not give quite the same improvement, however, but all the same this experiment showed that such a method could contribute to a richer and more complete search request.

Quantitative and qualitative studies of referential relations in unrestricted natural text are necessary both for a better theoretical understanding of referential processes, and for the development of empirically well-founded algorithms for anaphora resolution in the framework of natural language processing (NLP) systems. The aim of the study reported in this paper was to provide preliminary empirical data on anaphoric pronouns in Swedish. The relation between the pronoun and its antecedent was studied for 600 pronouns in three different types of unrestricted written Swedish text, and a simple pronoun resolution algorithm was tested on the sample.

In this paper we describe a multilevel model for natural language processing. The distinct computational strata are motivated by invariant linguistic properties which are progressively uncovered from utterances. We examine each level in detail. The processes are morphological analysis, dependency parsing, logico-semantic analysis and query adaptation. Both linguistic and computational aspects are discussed. In addition to theory, we consider certain engineering viewpoints important and discuss them briefly.

The article deals with the possibility of constructing a synopsis of a given text by automatic means from the text itself. It is suggested that this might be done on the basis of a statistical analysis of the vocabulary and collocations in the vocabulary, followed by simple parsing of selected parts of the text, and simple textual analysis, everything based on the results of the vocabulary analysis. The sentences and sentence fragments that are found to be central can then be used for the reconstruction of a substantially abridged version of the original text. After a discussion of the background of the model, a sample analysis is shown.

The paper examines the concept of semantic subordination in terminology and lexical semantics. Although the two traditions differ in their view of language, their inventories of semantic relations are very similar. It is shown that the defining characteristics of the generic relationships from the principles of terminology correspond to most of the salient features of contrast sets. It is argued that the onomasiological dictionaries constitute a valuable collection of data on the semantic relationships in the general language. Examples are given from Dansk Begrebsordbog. It is claimed that the registration of contrast sets is one step towards an operational definition of word sense.

The paper presents Referent Grammar (RG), a version of generalized phrase structure grammar. RG uses descriptive labels for defective categories (categories tacking a constituent) instead of slash expressions and needs no null (empty, zero) categories. RG uses both functional and categorial representations and the grammar rules, written in the Prolog DCG formalism, relate these two levels. The functional representations of RG include referent variables (numbers) with noun phrases which makes it possible to keep track of the referents within the sentences and in the text. Relative clauses can be defined in a new way using these referents and the referent, enriched with grammatical features, can be used for controlling agreement. Substantial fragments of Swedish and English grammars have been implemented in RG and the print-out of a demo session shows how RG works in computerized analysis, generation and translation of Swedish and English sentences.

The notion of symbolizability is taken as the second requisite of computation (the first being ‘algorithmizability’), and it is shown that symbols, qua symbols, are not symbolizable. This has far-reaching consequences for the computational study of language and for AI-research in language understanding. The representation hypothesis is formulated, and its various assumptions and goals are examined. A research strategy for the computational study of natural language understanding is outlined.

The machine translation project Eurotra is described as a multilanguage modular translation system with 9 monolingual analysis modules, 72 bilingual transfer modules, and 9 monolingual synthesis modules. The analysis module for Danish is described as a three-step parser with structure generation rules for immediate constituent structure, syntactic structure, and semantic structure, and translation rules between them. The topological grammatical description of Danish proposed by Paul Diderichsen, is shown to be useful in building the parser for Danish, especially with respect to the interaction between empty slots and filled slot in the topological pattern. Lastly, the special problem with parsing and disambiguation of sentences that allow many pp attachments patterns is mentioned and a solution is suggested.