I'm wondering how an MLT query with multiple "like" free text documents works. I assumed that it would basically either concatenate all the texts together, extract the best N words, and match some percent of them to get the results OR it would do that process separate for each like document and only return results that match to every like document.
However, I tried doing a basic test where I did the following:
- like: "this is my long text"
- like: ["this is my long text", "this is my long text"]
Basically I had two scenarios. The first one, I had some long text that I used as my like document. The second, I put the exact same long text in twice. I expected to return the same results in both cases. However, in the first case, I returned 8,271 results in my example. In the second case, I returned only 2 results.
Can anybody explain why that is?