Please enter your query. There is no query language element but comma ',' as separator that forces the query analysis to treat element separated by comma as distinct entities. The query evaluation recognizes multipart entities with patterns and treats them as single entities.
The query evaluation result consists of 3 parts. The first part returns the entities with closest vector (cosine similarity) in the collection. The collection knows of about 10 million named entities with an associated vector calculated with word2vec.
The second part returns the most popular links near query entities. Of course the search does not scan the whole collection. It extracts the links from the top 120 documents matching a query.
The third part is a ranklist calculated with BM25pff, a schema derived from BM25, taking proximity into account plus the link popularity depending on the query as calculated in part 2.
Enjoy your experience with the deep search of Strus on Wikipedia!