| Available from UMI in association with The British Library. Requires signed TDF.; This thesis concerns a new method of natural language processing, namely example or analogy-based processing. The example-based paradigm is compared and contrasted to, what might be called, conventional rule-based processing, and the advantages and disadvantages of both approaches are discussed.; One of the main themes of this thesis is how best to represent examples in an example-based environment such that some generative capacity can be afforded a system without recourse to rule-based techniques, thereby reducing the inflexibility associated with analogically-driven linguistic computation. The procedures of Cloning and Recombination have been developed as a consequence.; It is suggested that in the field of analogy-based language processing, researchers assume, to some extent at least, that conventional rule-based techniques will or should be used at some point in the architecture of such systems. This thesis attempts to realize a completely analogy-based architecture where rule-based processing is not used at all.; The importance of the description of examples is discussed with the introduction of a representative function which incorporates both linguistic and non-linguistic knowledge with the specific aim of representing examples in more than one language in order to facilitate translation between languages. The use of Functional Grammar and rhetorical text description are also used and discussed in relation to this goal.; Possibly the most important task in example-based processing is the determination of the similarity (or distance) between an input language fragment and a set of examples. Analogical Modelling is suggested as an interesting and useful mechanism with which to measure this distance, and a number of experiments are described which attempt to evaluate its effectiveness in this respect. |
