[
https://issues.apache.org/jira/browse/SOLR-17319?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18018131#comment-18018131
]
David Smiley commented on SOLR-17319:
-------------------------------------
To those not paying attention -- Sonu has a PR and lots of peer review.
Through subclass extension and some additional hooks in QueryComponent, there's
no negative/distracting impact to existing use-cases not using this opt-in
subclass of QueryComponent. We reviewers have been into the details but let's
consider the big picture of this PR (something I wish I did immediately as a
reviewer). The RRF algorithm in Sonu's PR is fundamentally only applied at
each shard for it's contribution to the whole. The cross-shard merge is
basically just an interleaving (not obvious looking at the code). That's very
unfortunate since many shards will degenerate to simply interleaving and not
RRF. I wish there was an initial design discussion before code / sunk-cost
investment. RRF of the whole is the ideal. Ideally sharding would have no
impact on the final results, notwithstanding per-shard term statistics (and
there's the obscurely named ExactStatsCache to handle that). A wholistic RRF
could be implemented at a later date but I suspect that would amount to a
rewrite of this implementation. That kind of solution would mean doing
distributed-search twice (once per sub-query) from a SearchComponent and then
doing RRF _over that_. It really wouldn't look much like the current code.
> Introduce support for Reciprocal Rank Fusion (combining queries)
> ----------------------------------------------------------------
>
> Key: SOLR-17319
> URL: https://issues.apache.org/jira/browse/SOLR-17319
> Project: Solr
> Issue Type: New Feature
> Components: vector-search
> Affects Versions: 9.6.1
> Reporter: Alessandro Benedetti
> Assignee: Alessandro Benedetti
> Priority: Major
> Labels: pull-request-available
> Time Spent: 22.5h
> Remaining Estimate: 0h
>
> Reciprocal Rank Fusion (RRF) is an algorithm that takes in input multiple
> ranked lists to produce a unified result set.
> Examples of use cases where RRF can be used include hybrid search and
> multiple Knn vector queries executed concurrently.
> RRF is based on the concept of reciprocal rank, which is the inverse of the
> rank of a document in a ranked list of search results.
> The combination of search results happens taking into account the position of
> the items in the original rankings, and giving higher score to items that
> are ranked higher in multiple lists. RRF was introduced the first time by
> Cormack et al. in [1].
> The syntax proposed:
> JSON Request
> {code:json}
> {
> "queries": {
> "lexical1": {
> "lucene": {
> "query": "id:(10^=2 OR 2^=1 OR 4^=0.5)"
> }
> },
> "lexical2": {
> "lucene": {
> "query": "id:(2^=2 OR 4^=1 OR 3^=0.5)"
> }
> }
> },
> "limit": 10,
> "fields": "[id,score]",
> "params": {
> "combiner": true,
> "combiner.upTo": 5,
> "facet": true,
> "facet.field": "id",
> "facet.mincount": 1
> }
> }
> {code}
> [1] Cormack, Gordon V. et al. “Reciprocal rank fusion outperforms condorcet
> and individual rank learning methods.” Proceedings of the 32nd international
> ACM SIGIR conference on Research and development in information retrieval
> (2009)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
