There’s several moving pieces to Fusion’s signals processing, so we’d certainly
encourage you to try Fusion out yourself and not reinvent these wheels :)
But here’s the general gist of how it works:
- events/signals/clicks are logged (to a separate Solr collection)
- periodically jobs run that aggregate signals into boostable weights, and
clean up used/unneeded/old signals
aggregated signals are stored in a separate Solr collection also
- query pipeline includes stages that look up signals matching the query,
prune/distill the retrieved signals,
and then incorporate them into the actual search request
Signals, in Fusion, are general purpose and could enable not only click
boosting but also search history boosting, item/item, query/item, and
item/query facilities.
—
Erik Hatcher, Senior Solutions Architect
http://www.lucidworks.com <http://www.lucidworks.com/>
> On Feb 5, 2016, at 5:48 AM, John Blythe <j...@curvolabs.com> wrote:
>
> hi all,
>
> i'm trying to find more information online about how to implement something
> similar to the 'signals' feature found in Fusion. so far i've found one
> decent article that isn't discussing the Fusion feature specifically. does
> any of you happen to have some solid resources to point me in the direction
> of concerning this?
>
> thanks-
