| With the rapid development of the construction industry, competition among enterprises becoming more serious, the main focus lies in the problem of construction bids, and whether success or defeat is decided by the construction costs. At present, most businesses analyze costs of building works by artificial methods, which is inefficient and high error rate. The main factors that determine project cost is the project acreage and the materials used. To improve enterprise efficiency, the natural language understanding technology is used to obtain more accurate project cost in this paper through the text analysis of related building structures and materials in construction maps.The text mainly expound about the status of natural language researching and the meaning of field-based limited natural language, analyzed the knowledge representation in knowledge-base, and established the knowledge-base based on the field of architecture; It analyzed the word auto-segmentation methods, designed auto-segmentation system according to the sub-word theory, and achieved the automatic- segmentation of construction texts, meanwhile, resolved the existing problems in the process of auto-segmentation. It introduced the semantic analysis and syntactic analysis methods; designed of the limited natural language understanding system, achieved the analysis and understanding of construction texts, and obtained engineering cost information according to the material value. |