We are building an enterprise search engine for about 130,000 documents. It includes FAQs, Small articles to 3-4 page documents, all of them are stored in structured JSON. We were originally planning to use Amazon Kendra to do the same, but unfortunately it was not approved by our enterprise cloud team.
Now we have to move to elasticsearch, looking to experiment several ways to improve search results with the use of machine learning/nlp.
Ex: search query correction with model trained using previous search data and it's results, generate alternative queries with same meaning to enhance the result etc.. Looking for ideas from here to experiment with