An automatic machine translation system for English-Maltese

Abstract

Machine Translation is a research field in Computer Science whose aim is to automate the process of translating text from one language to another using computers. This field is of social, political, commercial, scientific, intellectual and even philosophical interest. Research in MT has been going on since the 1940s, but unfortunately, good results have been difficult to obtain due to the inherent difficulty of the translation task. Briefly specified, an Automatic Machine Translation system operates by first parsing text in a source language into some kind of representation, and then using that representation to generate the corresponding target language text. Of course, this process must preserve the meaning, and possibly some of the style, of the original text. The kind of representation used influences heavily the general design of the system. While there is no recognised standard form of representation or generation technique, two distinct approaches can be described in the abstract. The first and most easily comprehensible is the interlingual approach, in which the representation is a natural language free meaning representation, that captures only the semantic content of the input sentence. This structure is then used as the basis for target language synthesis. The second kind of design is the transfer approach in which the representation structure remains much closer to the source language and still retains some of its linguistic characteristics. Transfer rules are used to transform this into another representation structure conveying the same meaning but containing the linguistic characteristics of the target language. A final phase translates this last structure into the output text.Such a specification of the translation process is highly over-simplified, and belies many details and problems that have to be overcome in order to achieve reasonable quality in translation output. Some of the problems, such as the syntactic and semantic analysis of texts, are ones that exist in the Natural Language Processing field of research. Others are specific to Machine Translation. This report documents a project that aimed to implement a Machine Translation system for translating between English and Maltese. The implementation of a general system capable of translating any arbitrary text is a task of huge proportions, so the translation was to be based on a text archive that would have to be created specifically for the purpose. This project was carried out by the undersigned, Paulseph-John Farrugia, in partial fulfilment of the requirements for the degree of Bachelor of Science (Hons). In order to provide a context for the main documentation of the project, various aspects of the field of Machine Translation are discussed first. These illustrate the history of this field, the issues involved, the results obtained from years of research, as well as the current state of the art. Describing in detail each of these aspects would require a report of huge proportions, so only the major ideas are given. A comprehensive literature list indicates where to refer to for further information.