bcn Algorithms for Linguistic Processing

Bibliography 55 Introduction Algorithms for Linguistic Processing is a research proposal in the area of computational linguistics. The proposal focuses on problems of ambiguity and processing efficiency by investigating grammar approximation and grammar specialization techniques. In this section we will try to explain for a broader audience what computational linguistics is, and what the current proposal is all about. What is computational linguistics? Linguistics is concerned with the study of natural languages such as Dutch, English, Korean, Nepali, Tagalog etc. Linguists hypothesise models ('gram-mars') of such languages which indicate the form utterances in the language can have. A grammar of English, for instance, will account for the fact that the first example below is an English sentence, whereas the second is not, even though it is made up of the very same English words: (1) a. The small man loves the tall lady b. The lady tall man loves small the c. The tall lady loves the small man Such grammars also describe the relation between form and meaning of the utterances in a given natural language. The grammar must distinguish the meaning of sentence (1-a) from the meaning of sentence (1-c). For the non symmetric verb love the difference in syntactic order corresponds not only to a difference in grammatical function (such as subject, direct object), but also to a difference in semantic roles (e.g. who is loved, and who is loving). Syntactic analysis will therefore determine the semantic interpretation. In linguistics, grammars can play various roles. For example in theoretical linguistics they are sometimes considered as models of the knowledge native speakers have of their language. In computational linguistics, another perspective is taken: computational linguistics is about the computation of the relation between form and meaning in natural language. Thus, rather than focusing on what the relation between form and meaning is, computational linguists want to know how this relation can be computed. Suppose we have a grammar 6 Gertjan van Noord of Dutch and we are given a Dutch sentence, how can we compute its corresponding meaning? And in the reverse direction, suppose we are given a certain meaning, how can we compute a Dutch or an English sentence expressing that meaning? Thus, computational linguistics is concerned with the way in which natural language is or can be processed. Why do we want to know how the relation between form and meaning can be computed? We identify …