Modeling an ERD can be done manually, but generally obtaining an Entity Relationship (ER) Diagram modeling manually will usually take a long time. So, it takes an ERD generator automation from the requirements specification to make it easier to do ERD modeling. This study will develop a system that produces ERD from requirements specifications in Indonesian by applying several stages of Natural Language Processing (NLP) according to needs research. The requirements specification used by the research team used technical document analysis. The stages of NLP used by the research team are: case folding, sentence segmentation, tokenization, POS tagging, chunking and parsing. Then the research team will conduct the words from the text that have been studied in the stages of NLP with the Rule-Based method to find a list of words that meet the ERD components such as: entities, attributes, primary keys and relations. The research team will describe the results obtained in the previous stage using the Graphviz library. From the results of the evaluation of the ERD system design, the research team used an expert evaluation evaluation. From the evaluation results obtained based on the evaluation of several cases, the results of the average precision, recall, and F1 scores from each expert are: 91%, 90%, 90% in expert 1; in expert 2 obtained 90%, 90%, 90%; in expert 3 obtained 98%, 94%, 96%; in expert 4 obtained 93%, 93%, 93%; and in expert 5 obtained 98%, 83%, 90%.
Copyrights © 2021