This research paper presents a new software machine for generating Thai sentences according to the objectives of communication: declarative sentences, negative sentences, question sentences, and command or request sentences. Then, those sentences will be stored in a corpus of Thai sentences for use in Text Generation and Story Generation research or Machine Learning and building a language model. The inner mechanisms of machine relies on relational databases using a cross-product relational algebraic function as a rule to control sentence construction followed by Thai grammatical patterns. The preliminary theoretical experiments imported 14 Thai sentence patterns and 4.506x10¹⁶ sentences with 8 Thai sentence patterns of the objective communication. The results of the theoretical experiments found that the machine was able to create a context-free sentence of 1.09928x10¹⁸, divided into the number of questions 6.41245x10¹⁷, negative sentences 2.74819x10¹⁷, and command or request sentences 1.83213x10¹⁷, respectively.