Implementation and Evaluation of an LFG-based Parser for Wolof

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 5128–5136 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 5128 Implementation and Evaluation of an LFG-based Parser for Wolof Cheikh Bamba Dione University of Bergen Sydnesplassen 7, 5007 Bergen dione.bamba@uib.no Abstract This paper reports on a parsing system for Wolof based on the LFG formalism. The parser covers core constructions of Wolof, including noun classes, cleft, copula, causative and applicative sentences. It also deals with several types of coordination, including same constituent coordination, asymmetric and asyndetic coordination. The system uses a cascade of finite-state transducers for word tokenization and morphological analysis as well as various lexicons. In addition, robust parsing techniques, including fragmenting and skimming, are used to optimize grammar coverage. Parsing coverage is evaluated by running test-suites of naturally occurring Wolof sentences through the parser. The evaluation of parsing coverage reveals that 72.72% of the test sentences receive full parses; 27.27% receive partial parses. To measure accuracy, the parsed sentences are disambiguated manually using an incremental parsebanking approach based on discriminants. The evaluation of parsing quality reveals that the parser achieves 67.2% recall, 92.8% precision and an f-score of 77.9%. Keywords: Parser, LFG, low-resource language, treebank, Wolof. 1. Introduction Deep grammars that follow an established linguistic theory such as Lexical Functional Grammar (LFG) (Bresnan, 2001) provide detailed syntactic analysis that is essential for the further development of NLP applications. This paper reports on the development and evaluation of the first LFG parsing system for Wolof, a low-resource Niger-Congo language mostly spoken in Senegal. The system is developed as part of the Parallel Grammar (ParGram) project (Butt et al., 2002) and is based on the Xerox Linguistic Environment (XLE) (Crouch et al., 2019). The goal of this research work is to provide a practical parsing system with broad coverage and deep analysis of naturally occurring Wolof data. The system described in this paper is the first parser reported for Wolof. LFG assumes two core levels of syntactic analysis: a c(onstituent)-structure which characterizes the phrase struc- ture configurations as a phrase structure tree, and a f(unctional)-structure which encodes grammatical relations (e.g. subject, object) and features (e.g. person, number). For instance, when coupled with XLE, the Wolof grammar assigns to example (1) the c- and f-structure in Figure 1.1 (1) Janq girl y-i NC-P bind write na-ñu FIN-3PL téeré book b-i. NC-P “The girls wrote the book.” The c-structure organization proposed for Wolof is briefly discussed in section 2.2.. The f-structure in Figure 1 indi- cates that the main predicate of (1) is bind ‘write’ and has a subject (SUBJ) and an object (OBJ). SUBJ has a seman- tic predicate (janq) and is analyzed as a noun that belongs to the b and y noun classes (see section 2.1.). The SPEC feature in the SUBJ f-structure is introduced by the definite (def) determiner yi, which has a semantic predicate (yi) and encodes deixis information (proximal). OBJ shows a similar f-structure. The sentence is an …

Content truncated to 500 words