Corpus Annotation within the French FrameNet: a Domain-by-domain Methodology

Corpus annotation within the French FrameNet: a domain-by-domain methodology Marianne Djemaa⋆, Marie Candito⋆, Philippe Muller⋄, Laure Vieu△ ⋆ Alpage (Univ. Paris Diderot / INRIA), ⋄ IRIT, Toulouse University , △ IRIT, CNRS, asfalda@inria.fr Abstract This paper reports on the development of a French FrameNet, within the ASFALDA project. While the first phase of the project focused on the development of a French set of frames and corresponding lexicon (Candito et al., 2014), this paper concentrates on the subsequent corpus annotation phase, which focused on four notional domains (commercial transactions, cognitive stances, causality and verbal communication). Given full coverage is not reachable for a relatively “new” FrameNet project, we advocate that focusing on specific notional domains allowed us to obtain full lexical coverage for the frames of these domains, while partially reflecting word sense ambiguities. Furthermore, as frames and roles were annotated on two French Treebanks (the French Treebank (Abeill´e and Barrier, 2004) and the Sequoia Treebank (Candito and Seddah, 2012), we were able to extract a syntactico-semantic lexicon from the annotated frames. In the resource’s current status, there are 98 frames, 662 frame-evoking words, 872 senses, and about 13000 annotated frames, with their semantic roles assigned to portions of text. The French FrameNet is freely available at alpage.inria.fr/asfalda. Keywords: FrameNet, French, semantic roles, semantic frames, semantically-annotated corpus 1. Introduction The ASFALDA project1 aims to build semantic resources and a corresponding semantic analyzer for French, to capture generalizations both over predicates and over the semantic arguments of predicates. We chose to build on the work resulting from the FrameNet project (Baker et al., 1998), which provides a structured set of prototypical situations, called frames, along with a semantic character- ization of the participants of these situations (called frame elements, but we’ll use roles for short). The corresponding English lexicon associates frames with the words that can evoke them (called frame-evoking elements, FEEs for short). While other English semantic resources, such as PropBank (Palmer et al., 2005) or VerbNet (Schuler, 2005), also provide semantic classes and/or semantic roles for predicate arguments, we chose FrameNet mainly because of its more semantic orientation, which is crucial for portability to other languages. FrameNet offers gener- alization over not only syntactic variation (e.g. diathesis alternation) but also lexical variation (like VerbNet but unlike PropBank), and groups together lexical units of various categories, on the basis of criteria that are not primarily syntactic (unlike VerbNet). The resources built within ASFALDA consist of a set of frames, a French lexicon in which lexical units are associated to FrameNet frames, and a semantic annotation layer added on top of existing syntactic French treebanks. The project also aims to investigate new models for frame-based semantic analysis. In a first phase of the project (Candito et al., 2014), the work focused on a set of notional domains, and frames pertaining to these domains were selected from the frame set of Berkeley FrameNet 1.5. These frames were adapted to French, and the corresponding French lexicon was built. The current 1alpage.inria.fr/asfalda paper focuses …

Content truncated to 500 words