can anyone explain how mlt works with the mix of like text and document Ids? We want to find similar documents of an selected document and the text of the query which the user has used. Following our workflow:
- User search documents by an fulltext query
- based on the result of the query of step 1 the user select one document as source of the more like this query
- execute mlt query with the id of the source document and the text of the fulltext query
If we provide both, the text of the fulltext query and the id of the source doucment, the result of the mlt query contains much more documents which doesn't match fine to our source document.
If we only provide the id of the source document and not the text of the fulltext query the result of the mlt query looks quite better. The goal which we want to achieve with the mix of text and document ids is a better ranking of documents which also matches to the text of the fulltext query.
Any hints to achieve this?