# Compute term frequency diff

**URL:** <https://discuss.elastic.co/t/compute-term-frequency-diff/193381>\
**Category:** Elasticsearch\
**Created:** [August 1, 2019, 5:18pm UTC](https://discuss.elastic.co/t/compute-term-frequency-diff/193381 "2019-08-01T17:18:55Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![aczapszys](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/aczapszys/32/51454_2.png) [@aczapszys](https://discuss.elastic.co/u/aczapszys)\
**Post date:** [August 1, 2019, 5:18pm UTC](https://discuss.elastic.co/t/compute-term-frequency-diff/193381/1 "2019-08-01T17:18:55Z")

</div>

I have an use case where I am comparing input titles to a corpus of titles to determine whether I have seen this title before. Using the default scoring match on the title field works great for ranking, but I also want to extract a comparison of the input title and the suggested title.

My thinking is that I can add a script\_fields to compute a TF difference as follows, but I'm in the weeds on this one. I'm not sure how to start. Can I script this as a field?

Pseudo-code:  
get terms from corpus title "corpus\_terms"  
get terms from input title "input\_terms"

```
let all_terms = union(corpus_terms , input_terms)
let total = 0
let total_pcts = 0
for term in all_terms:
    let weight = 0.0
    if term in input_terms:
        if term in corpus_terms:
            weight = 1.0
        else:
            weight = 0.0
    else:
        # Slight penalty for terms in corpus title not appearing in input_terms
        weight = -0.001

    let pct_of_titles_with_term = index[term].doc_count() / index.doc_count()
    total_pcts += pct_of_titles_with_term
    total += (pct_of_titles_with_term * weight)

let difference_from_0_to_1 = total / total_pcts
return difference_from_0_to_1
```

---

<div class="post-metadata">

**Author:** ![system](https://us1.discourse-cdn.com/elastic/original/3X/1/a/1ac57faf039f6b580b3f104ef42a2a89e41014de.png) [@system](https://discuss.elastic.co/u/system)\
**Post date:** [August 29, 2019, 5:19pm UTC](https://discuss.elastic.co/t/compute-term-frequency-diff/193381/2 "2019-08-29T17:19:08Z")

</div>

Closed on August 29, 2019, 5:19pm UTC

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