7. Conclusion

The work presented in this document comes from a growing necessity for providing complete information to judges and legal authorities. Formulating a decision or gather insight on legal domain is far from being trivial. This work provided insights on how to create a system that utilises semantic search through embeddings, exploring the immensely potential of neural networks, instead of using more traditional lexical techniques.
We explored the capability of ElasticSearch to host a semantic search system, developing the
necessary language models along the way. We handled different problems when treating raw text data, tackling the need to pre-process the text before performing any further tasks. Through working together with multiple Project IRIS members, we incorporate our solution within the main solution developed for STJ, and provided support for other related tasks.

This thesis, being a segment of a bigger project: Project IRIS, allowed to developed multiple soft and hard skills, by interacting with outstanding professionals and doctorate students. It provided insight on how to effectively research on a daily basis and how to co-operate in a large project with multiple members.