The Acquisition of Word Order by a Computational Learning System

The purpose of this work is to investigate the process of grammatical acquisition from data. We are using a computational learning systern that is composed of a Universal Grammar with associated parameters, and a learning algorithm, following the Principles and Parameters Theory. The Universal Grammar is implemented as a Unification-Based Generalised Categorial Grammar, embedded in a default inheritance network of lexical types. The learning algorithm receives input from a corpus annotated with logical forms and sets the parameters based on this input. This framework is used as basis to investigate several aspects of language acquisition. In this paper we are concentrating on the acquisition of word order for different learners. The results obtained show the different learners having a similar performance and converging towards the target grammar given the input data available, regardless of their starting points. It also shows how the amount of noise present in the input data affects the speed of convergence of the learners towards the target. 1 I n t r o d u c t i o n In trying to solve the question of how to get a machine to automatically learn linguistic information from data, we can look at the way people do it. Gold (1967) when investigating language identification in the limit, obtained results that implied that natural languages could not be learned only on the basis of positive evidence. These results were used as a confirmation for the proposal that children must have some innate knowledge about language, the Universal Grammar (UG), to help them overcome the problem of the poverty of the stimulus and acquire a grammar on the basis of positive evidence only. According to Chomsky's Principles and Parameters Theory (Chomsky 1981), the UG is composed of principles and parameters, and the process of learning a language is regarded as the setting of values of a number of parameters, given exposure to this particular language. We employ this idea in the learning framework implemented. In this work we are interested in investigating the acquisition of grammatical knowledge from data, focusing on the acquisition of word order, that reflects the underlying order in which constituents occur in different languages (e.g. SVO and SOV languages). The learning system is equipped with a UG and associated parameters, encoded as a Unification-Based Generalised Categorial Grammar, and a learning algorithm that fixes the values of the parameters to a particular language. The learning algor i thm follows the Bayesian Incremental Parameter Setting (BIPS) algorithm (Briscoe 1999), and when setting the parameters it uses a Minimum Description Length (MDL) style bias to choose the most probable grammar that describes the data well, given the goal of converging to the target grammar. In section 2 we describe the components of the learning system. In section 3, we investigate the acquisition of word order within this framework and discuss the results obtained by different learners. Finally we present some conclusions and future work. 2 The Learning System The learning system is composed of a language learner equipped with a UG and a learning algorithm that updates the initial parameter settings, based on exposure to a corpus of utterances. Each of these components is discussed in