The Prototype Development of Thai Language Understanding based on Mental Image Directed Semantic Theory

Nowadays, Artificial Intelligence (AI) has been embedded in various tools such as smart phones, even in those we might use unconsciously every day. However, AI technologies for automatic understanding human languages, namely, Natural Language Understanding (NLU) has been still far from its goal because of very difficult problems of semantics. This paper describes a prototype NLU system for Thai language based on Mental Image Directed Semantic Theory in order to help Thai language beginners more comprehensively. For the first step, Thai sentences of basic structures intended for amateur users were chosen for the system to process. In the system, input sentences are fragmented to detect constituent words and their grammatical relationships, and then transduced to semantic representations in a formal language named Language for Mental-image Description. The semantic interpretations are evaluated through reasoning processes such as disambiguation and entailment. Finally, the system returns animations as its understanding results of the inputs, in order to help language learners understand the text contents in a more intuitive way.