Hi, What Mikhail mentioned is the answer to all my questions but sadly it is not released yet. Is there any way I can use this unreleased version for now or when it is going to be released ?
and yes I want to rescore all topK results based on recency. Regards, Iram Tariq | Software Architect NorthBay Direct: +1 (902) 329-7329 iram.ta...@northbaysolutions.net www.northbaysolutions.com On Tue, Jan 2, 2024 at 5:50 AM Alessandro Benedetti <a.benede...@sease.io> wrote: > Hi Iram, following up on Mikhail's answer: > > 1) K Nearest Neighbour is a retrieval approach intended to look for the > closest(approximate) K vectors to a query one. > "override the existing function or write a > custom method to give a high score to the latest documents" > It seems suspicious to me. > It's like you want to combine two features for the final score: > > - Vector Similarity > - Recency > > As you can imagine a first question arises: > How do you want to combine these features? > Linearly? Non linearly? Do you want to re-score the top-k calculating this > recency? > Learning To Rank ( > https://solr.apache.org/guide/solr/latest/query-guide/learning-to-rank.html > ) > or general reranking( > https://solr.apache.org/guide/solr/latest/query-guide/query-re-ranking.html > ) > could be what you want. > Please make sure you are familiar with function queries as well ( > https://solr.apache.org/guide/solr/latest/query-guide/function-queries.html > ) > > 2) The pull request mentioned by Mikhail is on the spot, but it has not > been ported to Apache Solr yet. > Technically, it is going to be useful, but from a pragmatic perspective, I > am much more sceptical: > With current models, finding a threshold won't be easy at all. > > Cheers > -------------------------- > *Alessandro Benedetti* > Director @ Sease Ltd. > *Apache Lucene/Solr Committer* > *Apache Solr PMC Member* > > e-mail: a.benede...@sease.io > > > *Sease* - Information Retrieval Applied > Consulting | Training | Open Source > > Website: Sease.io <http://sease.io/> > LinkedIn <https://linkedin.com/company/sease-ltd> | Twitter > <https://twitter.com/seaseltd> | Youtube > <https://www.youtube.com/channel/UCDx86ZKLYNpI3gzMercM7BQ> | Github > <https://github.com/seaseltd> > > > On Mon, 25 Dec 2023 at 22:55, Iram Tariq <iram.ta...@northbaysolutions.net > > > wrote: > > > Hi All, > > > > Right now I am using the cosine similarity function for dense vectors. > > Is there any way I can override the existing function or write a > > custom method to give a high score to the latest documents. > > > > > > Also KNN Query Parser returns topK results matched with the input, but is > > there anyway possible we can get all documents for which similarity score > > is greater than a specific number? > > > > Any sort of answer will be helpful. Looking forward for the feedback. > > > > Regards, > > > > > > Iram Tariq | Software Architect > > > > NorthBay > > > > Direct: +1 (902) 329-7329 > > > > iram.ta...@northbaysolutions.net > > > > www.northbaysolutions.com > > >
