MWEs in Treebanks: From Survey to Guidelines

MWEs in Treebanks: From Survey to Guidelines Victoria Rosén1, Koenraad De Smedt1, Gyri Smørdal Losnegaard1, Eduard Bejček2, Agata Savary3 and Petya Osenova4 1University of Bergen, 2Charles University in Prague, 3François Rabelais University of Tours, 4Sofia University St. Kl. Ohridski and Bulgarian Academy of Sciences 1Bergen, 2Prague, 3Tours, 4Sofia victoria@uib.no, desmedt@uib.no, gyri.losnegaard@uib.no, bejcek@ufal.mff.cuni.cz, agata.savary@univ-tours.fr, petya@bultreebank.org Abstract By means of an online survey, we have investigated ways in which various types of multiword expressions are annotated in existing treebanks. The results indicate that there is considerable variation in treatments across treebanks and thereby also, to some extent, across languages and across theoretical frameworks. The comparison is focused on the annotation of light verb constructions and verbal idioms. The survey shows that the light verb constructions either get special annotations as such, or are treated as ordinary verbs, while VP idioms are handled through different strategies. Based on insights from our investigation, we propose some general guidelines for annotating multiword expressions in treebanks. The recommendations address the following application-based needs: distinguishing MWEs from similar but compositional constructions; searching distinct types of MWEs in treebanks; awareness of literal and nonliteral meanings; and normalization of the MWE representation. The cross-lingually and cross-theoretically focused survey is intended as an aid to accessing treebanks and an aid for further work in treebank annotation. Keywords: treebanks, MWEs, multiword expressions 1. Introduction PARSEME (PARSing and Multiword Expressions) is an in- terdisciplinary scientific network on the role of multiword expressions (MWEs) in parsing.1 Working Group 4 (WG4) in PARSEME is concerned with the enhancement of MWE- aware methodologies of treebank construction. A goal for the working group is to propose annotation guidelines for representing MWEs in treebanks. As a first step toward cre- ating such guidelines, WG4 has conducted a survey of exist- ing MWE annotations in treebanks (Rosén et al., 2015). The survey is open-ended and information on additional tree- banks is being added. Preliminary results of the survey show that there is considerable variation in the types of MWEs that are annotated in various treebanks, and also that the annotations for the same type of MWE can vary a lot de- pending on the language and the treebank type. Proposing detailed annotation guidelines is therefore a daunting task. As a first step towards recommendations, we propose some general principles for MWE annotations in treebanks. 2. The MWE Annotation Survey The survey was conducted by asking WG4 members with knowledge about particular treebanks to fill in some infor- mation in a wiki.2 The special purpose wiki was created in a Wikimedia-like framework, with a simple markup language and easy hyperlinking. It contains a table with a row for each treebank and columns for various types of MWEs, as shown in Figure 1. Each blue cell in the table is clickable and leads to an embedded information page. 1http://parseme.eu 2http://clarino.uib.no/iness/page?page-id=MWEs_ in_Parseme The MWE Types The table headers show the types of MWEs to be described. The typology chosen is syntactically based, and the types are among the most common ones …

Content truncated to 500 words