Tibor Kiss

Unsupervised Multilingual Sentence Boundary Detection

In this article, we present a language-independent, unsupervised approach to sentence boundary detection. It is based on the assumption that a large number of ambiguities in the determination of sentence boundaries can be eliminated once abbreviations have been identified. Instead of relying on orthographic clues, the proposed system is able to detect abbreviations with high accuracy using three criteria that only require information about the candidate type itself and are independent of context: Abbreviations can be defined as a very tight collocation consisting of a truncated word and a final period, abbreviations are usually short, and abbreviations sometimes contain internal periods. We also show the potential of collocational evidence for two other important subtasks of sentence boundary disambiguation, namely, the detection of initials and ordinal numbers. The proposed system has been tested extensively on eleven different languages and on different text genres. It achieves good results without any further amendments or language-specific resources. We evaluate its performance against three different baselines and compare it to other systems for sentence boundary detection proposed in the literature.

Local Modelling of Non-Local Dependencies in Syntax

Syntactic dependencies like movement, reflexivization, agreement, control, deletion, case assignment and binding can be non-local, crossing a clause boundary. A strategy pursued in many contemporary theories is to model cases of non-local dependencies in a local way, by successively passing on information in small syntactic domains. The present volume brings together eighteen articles that investigate the phenomenon from different theoretical backgrounds, based on evidence from a variety of typologically distinct languages.

Scaled log likelihood ratios for the detection of abbreviations in text corpora

We describe a language-independent, flexible, and accurate method for the detection of abbreviations in text corpora. It is based on the idea that an abbreviation can be viewed as a collocation, and can be identified by using methods for collocation detection such as the log likelihood ratio. Although the log likelihood ratio is known to show a good recall, its precision is poor. We employ scaling factors which lead to a strong improvement of precision. Experiments with English and German corpora show that abbreviations can be detected with high accuracy.

Produktivität und Idiomatizität von Präposition-Substantiv-Sequenzen

Abstract The present paper discusses two fundamental properties of so-called determinerless PPs, their idiomaticity and productivity. Combining evidence from German newspaper corpora with methods from statistical natural language processing, we can show that determinerless PPs are no more idiomatic that ordinary phrasal combinations in general. From an empirical point of view determinerless PPs can be classified as productive. Consequently determinerless PPs should not be treated as idiomatic exceptions, which can be covered by finite lists.

Integrating syntactic and prosodic information for the efficient detection of empty categories

We describe a number of experiments that demonstrate the usefulness of prosodic information for a processing module which parses spoken utterances with a feature-based grammar employing empty categories. We show that by requiring certain prosodic properties from those positions in the input, where the presence of an empty category has to be hypothesized, a derivation can be accomplished more efficiently. The approach has been implemented in the machine translation project VERBMOBIL and results in a significant reduction of the work-load for the parser.

Measuring the Productivity of Determinerless PPs

We determine the productivity of determinerless PPs in German quantitatively, restricting ourselves to the preposition unter. The study is based on two German newspaper corpora, comprising some 210 million words. The problematic construction, i.e. unter followed by a determinerless singular noun occurs some 16.000 times in the corpus. To clarify the empirical productivity of the construction, we apply a productivity measure developed by Baayen (2001) to the syntactic domain by making use of statistical models suggested in Evert (2004). We compare two different models and suggest a gradient descent search for parameter estimation. Our results show that the combination of unter+noun must in fact be characterized as productive, and hence that a syntactic treatment is required.

Annotating Spatial Interpretations of German Prepositions

Traditional approaches to prepositions classify preposition lexemes into different semantic classes like spatial, temporal or modal ones. In this paper we reconsider the so-called spatial prepositions and their sub classification. We propose a sense based classification of prepositions using traditional criteria to determine different classes of spatial-preposition interpretations. Our work on spatial preposition senses is part of a bigger project of developing an annotation scheme for a subset of German prepositions.

Reflexivity and Dependency

The analyses of Pollard and Sag (1992, 1994) and Reinhart and Reuland (1993) answer why certain reflexive pronouns – so-called picture-NP-reflexives – behave anomalously in allowing co-reference with nonbinding antecedents. These reflexives are exempt from Binding Theory’s Principle A because Principle A only requires binding if a potential antecedent is available in a given local domain. The reflexive himself in (1) does not find a local antecedent in the domain of the predicate picture and hence becomes exempt from an application of Binding Theory. Just like the co-indexation of the pronoun him with the subject John, the co-indexation of the reflexive is not an instance of binding, but an indication of co-reference.

Bedingungen für den Wegfall eines Artikels: Distribution und Interpretation von Präposition-Nomen-Kombinationen

Kombinationen aus Praposition und artikelloser Nominalprojektion, deren syntaktischer Kopf ein zahlbares Substantiv im Singular ist, fristeten lange Zeit ein Schattendasein in der Grammatikschreibung. Sie wurden ignoriert oder als Ausnahmen beschrieben, obwohl sie offenkundig regelhaft gebildet werden. Im vorliegenden Aufsatz verwenden wir computerlinguistische Verfahren, insbesondere „Annotation Mining“ und logistische Regression, um die syntaktische Distribution dieser Kombinationen zu charakterisieren und anhand zweier Prapositionen (‚ohne‘ und ‚unter‘) detailliert die Realisationsbedingungen zu bestimmen.

Erläuterungen zur Umsetzung einer HPSG im Basisformalismus STUF III

Der vorliegende Aufsatz beschreibt einige Grundzuge der Implementation einer HPSG im Verbmobil-Basisformalismus STUF III. Wir beginnen mit Betrachtungen zur Implementation der Grammatik und stellen einzelne Module der implementierten Grammatik vor. Hierbei diskutieren wir das Typsystem, das Konzept des Typmodules, die Behandlung von Sorten und ihre Beziehung zur Verarbeitung durch Regelinstanzen. Im vorliegenden Papier wird auf theoretische Untersuchungen zur Grammatik des Deutschen verzichtet; im Vordergrund steht vielmehr zu vermitteln, wie die einzelnen Konstrukte von STUF III bei der Grammatikimplementierung eingesetzt werden konnen. In the following paper, we describe some aspects of an implementation of an HPSG-style grammar in the Verbmobil Basic Formalism STUF III. The presentation includes basic design considerations as well as fundamental properties of the current STUF III grammar, including the type system, type modules, constraints, and rule instances. The purpose of the following paper is to give the reader some basic guidelines regarding the representation of linguistic information in STUF III. Hence, theoretical analyses or reflections on the grammar of German are omitted from the current paper.