Constructions and Frames - Advances in Frame Semantics, January 2011

In this paper, we present a frame semantic analysis of a small group of Italian verbs expressing visual perception, which constitutes the first stage of a project for developing an Italian FrameNet. Our results show a close correspondence between English and Italian perception-related frames. The main element of novelty in our approach is that the creation and annotation of Lexical Units is grounded in distributional information automatically acquired from a large, dependency-parsed corpus, which is balanced against the annotator’s linguistic intuition. We claim that this can help to overcome some of the shortcomings of the classical lexicographic method used to create FrameNet.

The FrameNet approach to text semantic annotation can be a reliable model to make the linguistic information and semantic content of legal texts explicit. This hypothesis is discussed and empirically demonstrated through a trial of annotating a corpus of Italian legal texts. This study aims to show that FrameNet is particularly appropriate to provide new perspectives for legal language studies and for legal knowledge representation tasks. Moreover, by relying on the output of a statistical dependency parser, the FrameNet-based annotation methodology presented here can be used successfully in the automatic semantic processing of legal texts.

It is not satisfactory to analyze temporal metaphors such Moving Ego (e.g. We are approaching the end of the year) and Moving Time (e.g. The end of the year is approaching) simply as mappings from SPACE to TIME. In this case, the intuitively obvious experiential basis of Moving Ego leads to a paradox if applied (with suitable adjustments) to Moving Time. Frame analysis makes possible an adequate account of the experiential basis of both metaphors.

Theoretically as well as empirically, paraphrase is a pivotal concept in many academic and nonacademic fields. And yet, its investigation has made very slow progress, due mainly to the lack of a framework that is versatile enough to deal with the nebulous nature of paraphrase in use. This paper demonstrates how the mechanisms of FrameNet can be utilized as a resource for systematic and coherent research into paraphrase. The semantic framework it provides, including detailed frame descriptions, frame-to-frame relations, and the recording of syntactic information, allows one to see in more principled ways why some set of sentences can be considered paraphrases of each other.