0
$\begingroup$

In our job portal search we are performing search by using dense embedding of Job Title + Skills across a pool of candidates. After this first step of retreival I'm planning to perform rerank based on number of skills matched. For example: 4 out of 7 skills matched

Since I'm using dense embedding I can't derive exactly how many skills matched and I can't simply do a keyword overlapping because that would fail for sceanrios like "ETL Tools" and "ETL", "Data Analysis" and "Data Visualisation" etc..

I'm planning to use SPLADE embedding to acheive this, is SPLADE a good way to solve this? or are there any other better ways of hanling this?

I can think of adding synonyms but that would be too difficult to maintain.

Improve this question
$\endgroup$

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.