This paper presents a novel approach to constructing multilingual lexical databases using semantic frames. Starting with the conceptual information contained in the English FrameNet database, we propose a corpus-based procedure for producing parallel lexicon fragments for Spanish, German, and Japanese, which mirror the English entries in breadth and depth. The resulting lexicon fragments are linked to each other via semantic frames, which function as interlingual representations. The resulting parallel FrameNets differ from other multilingual databases in three significant points: (1) They provide for each entry an exhaustive account of the semantic and syntactic combinatorial possibilities of each lexical unit; (2) They offer for each entry semantically annotated example sentences from large electronic corpora; (3) By employing semantic frames as interlingual representations, the parallel FrameNets make use of independently existing linguistic concepts that can be empirically verified. 1

Research in Construction Grammar assumes no strict separation between syntax and the lexicon. However, recent work by Goldberg (1995Goldberg ( , 2006 shows that there is indeed a separation between lexical entries and grammatical constructions, including constraints regulating the fusion of grammatical constructions with verbs. This paper argues that Goldberg's characterization of the interactions between lexical entries and grammatical constructions faces some of the same difficulties as the interactions between lexical entries and transformational rules in the Chomskyan framework (Chomsky, 1965(Chomsky, , 1981(Chomsky, , 1995. Drawing on a variety of corpus data this paper presents specific proposals that should be considered in order to arrive at a solution that overcomes difficulties inherent to Goldberg's approach. Based on a discussion of the concepts of analogy, collocational restrictions, frequency, and productivity this paper proposes to encode different types of semantic, pragmatic, and syntactic information in such a way that it is possible to account for a given utterance from a comprehension perspective, as well as a production perspective.

According to Goldberg, the caused-motion and resultative semantics associated with the constructions are encoded syntactically by a [NP V NP PP/AP] frame that typically does not occur with verbs such as those in (1) and (2). That is, in (2) the final interpretation of the combined verbal and constructional semantics is 'X CAUSES Y TO MOVE Z by drinking'. This meaning is paired with a three-argument syntactic frame [NP V NP PP] that expresses the combined semantics of the verb and the construction. One advantage of this approach lies in the fact that it is not necessary to posit implausible verb senses based on intransitive verbs in order

No abstract

No abstract