Natural Language to Unified Modeling Diagrams; A Deep Learning Approach

This paper proposes a suitable technique for the generation of Unified Modeling Diagrams (UML) from textual information using Natural Language Processing (NLP) and Deep learning concepts. The deep learning technologies can be used for the text analytics to improve and automate the UML diagram preparation. This technique initially focuses on the generation of consolidated design model from which the class model can be constructed. Here the class names, attributes, and associations are identified based on certain heuristic rules. Also the proposed method is evaluated and compared with certain known methods. In this paper generation of UML diagrams from textual information using NLP is discussed with the concepts of deep learning. The proposed algorithm is able to identify the required objects, attributes, relationships, etc., which will be specified by the user for the preparation of UML diagrams. This is possible by the various generation principles. Initially generation of consolidated design model is acquired and from that the class model is derived. Different features like class names, attributes, and associations are identified by heuristic rules. The proposed method is compared with conventional methods and it shows an improvement in accuracy from 4.1% to 8.7%.