> SELECT table_name, record_id, title, details
> FROM movie, actor, director
> WHERE any_field HAS TEXT SIMILAR TO 'usual suspects'
> ORDER BY relevance_factor DESC;

What about:

SELECT t.score, a.some, m.other, d.fields
FROM textSearch('actor,movie,director','suspects spacey') t
LEFT OUTER JOIN actor a ON t.record_id = a.record_id
LEFT OUTER JOIN movie m ON t.record_id = m.record_id
LEFT OUTER JOIN director d ON t.record_id = d.record_id
WHERE t.score > 0.8
ORDER BY 1

Rest of the discussion about the references between entites is
irrelevant to my opinion. We had a discussion on this topic approx.
one year ago and concluded that Google-like scoring has no use in
RDBMSes (at least that' my opinion from that discussion) since
information has completely different nature.

Roman