Objective

This research aims to leverage Natural Language Processing (NLP) techniques to advance the analysis of the Common Criteria security certification artefacts that are available as unstructured pdf documents with technical content. The primary goal is to design and implement a proof-of-concept solution that allows users to retrieve relevant segments within these certificates – a segment is defined as a coherent block of text within the PDF.

• Conduct a survey of existing approaches and select the most viable one considering the project's requirements.
• Develop a solution suitable for use as an information retrieval system. The system should be capable of highlighting segments most relevant to a given user query, ensuring a single PDF file is searched and the semantic meaning of the segments is considered.
• Evaluate the relevance of the retrieved segments using a predefined set of user queries, which will be compiled by the members of the CRoCS laboratory. Compare the proposed solution against a recognized baseline, such as bm25, to measure its efficacy.
• As a concluding part of the thesis, provide a comprehensive list of potential improvements and directions for future research.