Automatic Induction of Semantic Classes for German Verbs
Sabine Schulte · 2004
The verb is an especially relevant part of the sentence, since it is central to the structure and the meaning of the sentence: The verb determines the number and kind of the obligatory and facultative participants within the sentence, and the proposition of the sentence is defined by the structural and conceptual interaction between the verb and the sentence participants. For that reason, lexical verb information represents the core in supporting computational tasks in Natural Language Processing (NLP) such as lexicography, parsing, machine translation, and information retrieval, which depend on reliable language resources. But especially lexical semantic resources represent a bottleneck in NLP, and methods for the acquisition of large amounts of semantic knowledge with comparably little manual effort have gained importance. In this context, I am concerned with the potential and limits of creating a semantic knowledge base by automatic means, semantic classes for German verbs. Semantic verb classes generalise over verbs according to their semantic properties. They represent a practical means to capture large amounts of verb knowledge without defining the idiosyncratic details for each verb. The class labels refer to the common semantic properties of the verbs in a class at a general conceptual level, and the idiosyncratic lexical semantic properties of the verbs are either added to the class description or left underspecified. Examples for conceptual structures are Position verbs such as liegen ‘to lie’, sitzen ‘to sit’, stehen ‘to stand’, and Manner of Motion with a Vehicle verbs such as fahren ‘to drive’, flie gen ‘to fly’, rudern ‘to row’. A semantic classification demands a definition of semantic properties, but it is difficult to automatically induce semantic features from available resources, both with respect to lexical semantics and conceptual structure. Therefore, the construction of semantic classes typically benefits from a long-standing linguistic hypothesis which asserts a tight connection between the lexical meaning of a verb and its behaviour: To a certain extent, the lexical meaning of a verb determines its behaviour, particularly with respect to the choice of its arguments (Pinker, 1989; Levin, 1993). We can utilise this meaning-behaviour relationship in that we induce a verb classification on basis of verb features describing verb behaviour (which are easier to obtain automatically than semantic features) and expect the resulting behaviour-classification to agree with a semantic classification to a certain extent. A common approach to define verb behaviour is captured by the diathesis alternation of verbs. Alternations are alternative constructions at the syntax-semantic interface which express the same or a similar conceptual idea of a verb. In Example (1), the most common alternations for the Manner of Motion with a Vehicle The work reported here was performed while the author was a member of the DFG-funded PhD program ‘Graduiertenkolleg’