An Exploration of Connectionist and Symbolic Language Acquisition Techniques

The problem of language acquistion was originally studied by psychologists, but now it is an important problem in Artificial Intelligence as well. Rumelhart and McClelland (from now on referred to as R&M) have claimed that simulating the internal workings of the brain (i.e. using a connectionist method) best models language acquistion. This problem is important for several reasons. It has many practical applications. For example, one could imagine an advanced grammar check program for a word processor that would not need to have been programmed to handle a certain word in order to detect if its past tense was handled properly. In fact, such a program could theoretically deal with words that did not exist when it was programmed with relative ease, a feature that is becoming more important as new technologies and therefore verbs are being invented all the time. It is also possible that solving this problem would shed light on other fields, including Psychology and Linguistics. In this paper we present the history of the debate between Symbolic and Connectionist methods for language acquisition. In section 2, we describe R&M’s early work. In section 3 we examine objections to R&M by the proponents of Symbolic methods, as well as a response to these objections from Connectionists. In section 4, we present the first symbolic method for learning the past tense. In section 5 we give examples of more recent work in the field of language acquistion. In section 6, we look to possible future work.