Hi,
I’m reading about the ranking functions [1], and I have a couple questions…
1. Is ts_rank taking proximity of terms into account? It seems like it is, but the docs suggest that only ts_rank_cd does that.
2. Is there a way to search multiple terms like ‘a | b | c …’ but score higher when multiple match, AND take into account distance between words? It doesn’t seem like basic use of ts_rank or ts_rank_cd is doing this. Do you recommend a custom ranking function here?
For example, I want to search for “black bear” and get better results ordered so that documents with both words close together score highest, and the document with only “bear" is the last.
create table search_test ( title text, body text, vec tsvector );
— These 3 have “black” and “bear” at different distances from each other
insert into search_test values ('close', 'The black bear sat on a brown rock and ate a barrel of red berries.');
insert into search_test values ('medium', 'The brown bear sat on a black rock and ate a barrel of red berries.’);
insert into search_test values ('far', 'The brown bear sat on a red rock and ate a barrel of black berries.’);
— This one has the word “bear”, but not “black"
insert into search_test values ('only bear', 'The brown bear sat on a red rock and ate a barrel of orange berries.');
update search_test set vec = to_tsvector(body);
Now a query:
select title, ts_rank(vec, q) as rank
from search_test, to_tsquery('black & bear') q
where vec @@ q order by rank desc;
That surprises me by scoring close > medium > far. Hence, my question #1. Substituting ts_rank_cd also works, as expected.
If I change the query to `black | bear`, to try to match “only bear” as well, then both ts_rank and ts_rank_cd return equal rankings for “close”, “medium” and “far”.
Any recommendations?
thanks,
Rob
[1] http://www.postgresql.org/docs/9.4/static/textsearch-controls.html#TEXTSEARCH-RANKING
--
Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general
