Parsing a Polysynthetic Language

Proceedings of Recent Advances in Natural Language Processing, pages 562–567, Hissar, Bulgaria, 12-14 September 2011. Parsing a Polysynthetic Language Petr Homola Codesign, s.r.o. phomola@codesign.cz Abstract We present the results of a project of build- ing a lexical-functional grammar of Ay- mara, an Amerindian language. There was almost no research on Aymara in compu- tational linguistics to date. The goal of the project is two-fold: First, we want to provide a formal description of the lan- guage. Second, NLP resources (lexicon and grammar) are being developed that could be used in machine translation and other NLP tasks. The paper presents for- mal description of selected properties of Aymara which are uncommon in well- researched Western languages. Further- more, we present an experimental machine translation system into Spanish and En- glish. 1 Introduction Aymara is an Amerindian language spoken in Bo- livia, Chile and Peru by approx. two million peo- ple. It is a polysynthetic language that has many lexical and structural similarities with Quechua but the often suggested genetic relationship be- tween these languages is still disputed. The only research on Aymara in the field of computational linguistics we know about is the project described in (Beesley, 2006). The presented project uses Lexical-Functional Gram- mar (LFG) (Kaplan and Bresnan, 1982; Bresnan, 2001) to formally describe the lexicon, morphol- ogy and syntax of Aymara in a manner suitable for natural language processing (NLP). The gram- mar we have implemented is capable of parsing complex sentences with embedded clauses. We have also done experiments with machine transla- tion (MT) into Spanish and English; the results are presented in Section 4. Aymara is a polysynthetic language with a very complicated system of polypersonal agreement (see Section 2.3 for a brief description). A rare property of words in Aymara is the so-called vowel elision (sometimes called ‘subtractive morphol- ogy’) which is quite hard to describe formally. We show how vowel elision can be dealt with in the lexicon. The paper is organized as follows: Section 2 presents selected properties of Aymara, many of them absent from well-researched languages such as English, and their formal analysis in LFG. Section 3 introduces a dependency-based abstrac- tion of f-structures which brings formal grammars closer cross-linguistically. Section 4 describes our experiment with MT from Aymara into Span- ish and English. Finally, we conclude in Section 5 and give an outlook for further research. 2 Some Properties of Aymara In this section, we focus on some properties of Aymara at the level of morphology and syntax which are mostly absent from Western languages such as English, and sketch their analysis in LFG. A detailed description of the language can be found in (Hardman et al., 2001; Adelaar and Muysken, 2007; Cerrón-Palomino and Carvajal, 2009; Briggs, 1976). 2.1 Agglutinative Morphology Aymara has a very rich inflection. Suffixes of various categories can be chained to build up long words that would be expressed by a sentence in languages like English. For ex- ample, alanxarusksmawa (ala-ni-xaru-si-ka-sma- wa) means “I am preparing myself to …

Content truncated to 500 words