Wildcard search + thesaurus on xquery

Question

Hi All,
I just realized that thesaurus matching is not supported with wildcard search.
Here is current codes.
<Searches>
if (fn:exists(http:get-query-parameter("Ntt"))) then
for $paramNtt in http:get-query-parameter("Ntt")
     return <Search Mode="AllPartial" Key="i_products_st">{concat("*",concat($paramNtt,"*"))}</Search>
else ()
</Searches>
So, now i am thinking that I can change to somthing like following eql
<EqlExpression>
collection()/record[ endeca:matches(.,"i_products_st",$paramNtt,"AllPartial") or endeca:matches(.,"i_products_st",concat("*",concat($paramNtt,"*")),"AllPartial") ]
</EqlExpression>
though I didn't test it out yet. I guess there might be some other solution.
So, I would like to ask if any better way can solve this issue.
Thanks,

Answer 1

Are you trying to retrieve the union of (a) records that match the user's search term as literally entered, perhaps expanded by thesaurus and (b) records that match the user's search term as a substring via wildcard?
If so, then EQL can express that logic. For example, suppose that:
* the thesaurus contains an entry "truck" <-- --> "lorry"
* the user enters "truck" in the search box
EQL: collection()/record[ endeca:matches(.,"i_products_st","truck","AllPartial") or endeca:matches(.,"i_products_st","\*truck\*","AllPartial") ]
This search filter expression would match records containing: { truck, lorry, trucker, trucking, struck, ... }
This kind of infix wildcard expression has very broad recall. In this case you would probably also want to assert a relevance ranking strategy based on matching the user's original entry "truck", using Nrr.
Also worth thinking about search inputs that are multiple words like "truck driver" - would you want to implicitly wildcard every token separately e.g. \*truck\* \*driver\*
Finally, definitely worth thinking about search inputs that give pathological queries when implicitly wildcarded, like very short tokens e.g. a ==> \*a\*