ReForm: Automated Requirements Formalization

ReForm is a tool for the automated formalization of textual requirements using natural language processing (NLP) and machine learning (ML) techniques. It parses requirements written in natural language to extract the relevant information and automatically generate requirements models.

It utilizes various NLP methodologies such as dependency parsing and part-of-speech (POS) tagging as well as semantic role labeling (SRL) to extract model entities such as signals, operators, and parameters. The generated formalized requirements are represented in a human- and machine-readable format.

• Read textual requirements (e.g. CSV/JSON), analyze syntactic and semantic features, and generate a model representation (e.g. temporal logic, given-when-then structure, UML sequence diagrams)
• Optionally read textual signal/event descriptions (e.g. CSV/JSON), calculate similarity scores to requirements phrases and use them for better model prediction
• Easy adaptation of the algorithm's mapping rules to new use cases and documents
• Visualization and easy modification using via a web UI (streamlit)

Python, PyTorch, spaCy, AllenNLP, NLTK, Sentence-Transformers, Scikit-learn, NumPy, Pandas, (streamlit)

Requirements Engineers, Software Engineers

Open-source, licensing and permission required