The Morphological Analysis Of Bahasa Malaysia

This paper describes a model for the automated morphological analysis of Bahasa Malaysia (the Malay language) via the ATEF system, a component of the mechanical translation system known as ARIANE, which was developed by G.E.T.A. at Grenoble. This model serves two purposes, that is, to test the capability of handling Bahasa Malaysia morphological analysis using ATEF and also to provide a first working model. This grammar covers the three main morphological processes in Bahasa Malaysia analysis, that is, affixation, reduplication and compounding. Reduplication is the process whereby a base or some part of the base is repeated. There are three types of reduplication proper, rhyming and chiming. Reduplication of nouns generally gives a semantic category of heterogeneity or indefinite plural while reduplication of verbals result in one of the following semantical features: repetition, continuity, habituality, intensity, extensiveness and resemblance. Compound forms are constructions that have two or three free forms as their constituents and each of the constituent forms may either be a rootform or derived form. Affixation is a morphological process whereby a base may be extended by one or more affixes. Affixes may be classified as prefixe~ suffixes, infixes and circumfixes. Multiple affixation is also not uncommon in Bahasa Malaysia though no construction exceeds three layers of affixation. Several features are obtained through affixation. On affixation, morphographemic changes may occur depending on the initial segment of the rootform, word classes of the derived words are set and also semantical features are set. The setting of semantical features may be further complicated as a result of multiple affixation. Unlike affixation, the handling of reduplication and compounding do not present much of a problem for ATEF and is quite straightforward. Affixation is a more complicated process but also a more important process. A simple finite state diagram is used to depict the basic overall structure for the handling of multiple affixation as a more detailed finite state diagraw is not justifiable. Morphographemic changes occur mainly with the prefixes pe N and me N in which different allomorphs are ~sed depending on the initial segments of the derived word. On deletion of these allomorphs to obtain the resultant form, segments may have to be added to the resultant form, the form remains unchanged or substituted. This means that rules have to be provided for the treatment of each of these allomorphs individually. The main word classes in Bahasa Malaysia are nominals, verbals, auxiliaries, adverbals and particles and these again can be subcategorised. The word class of a derived word is dependent on its affix and on affix deletion to obtain the rootform, its word class is set. For multiple affixation, the outermost prefix (if any) determines the word class. Affixation also results in modifications or additional semantical features. Each affix carries a set of possible semantical features. For example, the prefix pe N may cause the wordform to be agentive, i~strumental, the object of action, etc. Which is the correct 'role' depends on the base on which the prefix was attached. In this model, no decision is made as to which semantical feature is the correct one. Instead, the whole set of features are set when the affix has been detected. Although it may be possible to subcategorise the word classes into groups with common semantical features, this model so far only considers grouping according to word classes and does not consider subgroupings for semantical features. Work is now in progress to include such subgrouping to provide a more complete morphological analysis of Bahasa Malaysia. This model not only handles these three main morphological processes, but also handles idiomatic expressions as well. On completion of morphological analysis, all the information gathered is submitted to the next stage of the ARIANE system, that is, the syntactical analysis stage in order to build up a more complete 'picture' of Bahasa Malaysia.