What is boost in similarity score function?


I have a question regarding the similarity function with BM25.

I understand tf and idf. However it's not clear for me what the boost variable is and what it depends on.

When I get the explanation of the score for a document I get the following:

"explanation": {
        "value": 13.391564,
        "description": "weight(text:despierto in 165046) [PerFieldSimilarity], result of:",
        "details": [
                "value": 13.391564,
                "description": "score(freq=3.0), computed as boost * idf * tf from:",
                "details": [
                        "value": 2.2,
                        "description": "boost",
                        "details": []

From what I have seen, boost is always a multiple of 2.2, but why is it sometimes 2.2, sometimes 4.4, or 6.6 etc?

Thank you in advance!

I have found it myself through trial and error.

Turns out the 'boost' is part of Okapi BM25

By default, k1=1.2, therefore, in the numerator of the equation we have f(q_i, D) * (k_1 + 1) = tf * 2.2. This is the boosting part.
The boost is simply tf of the query multiplied by (k_1+1). So, if a term appears once in the query, the boost will be just 2.2. However, if a term appears n times, it will be n*2.2.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.