yetangitu/recoll
1
0
Fork 0
Code Issues Pull requests Releases Wiki Activity

Raw dbs: do not skip diacritics and case expansion for terms which do not look like like natural language words (and for which stemming is skipped)

Browse source
This commit is contained in:
Jean-Francois Dockes 2012-11-16 13:15:58 +01:00
parent a29aea17ff
commit 494c57ee9c
1 changed files with 7 additions and 14 deletions
Download patch file Download diff file Expand all files Collapse all files

21
src/rcldb/expansiondbs.cpp
View file

@ -84,21 +84,7 @@ bool createExpansionDbs(Xapian::WritableDatabase& wdb,
for (Xapian::TermIterator it = wdb.allterms_begin(); for (Xapian::TermIterator it = wdb.allterms_begin();
it != wdb.allterms_end(); it++) { it != wdb.allterms_end(); it++) {
// Skip terms which don't look like natural language words.
if (!Db::isSpellingCandidate(*it)) {
LOGDEB1(("createExpansionDbs: skipped: [%s]\n", (*it).c_str()));
continue;
}
// Detect and skip CJK terms. // Detect and skip CJK terms.
// We're still sending all other multibyte utf-8 chars to
// the stemmer, which is not too well defined for
// xapian<1.0 (very obsolete now), but seems to work
// anyway. There shouldn't be too many in any case because
// accents are stripped at this point.
// The effect of stripping accents on stemming is not good,
// (e.g: in french partimes -> partim, parti^mes -> part)
// but fixing the issue would be complicated.
Utf8Iter utfit(*it); Utf8Iter utfit(*it);
if (TextSplit::isCJK(*utfit)) { if (TextSplit::isCJK(*utfit)) {
// LOGDEB(("stemskipped: Skipping CJK\n")); // LOGDEB(("stemskipped: Skipping CJK\n"));
@ -117,6 +103,13 @@ bool createExpansionDbs(Xapian::WritableDatabase& wdb,
} }
#endif #endif
// Dont' apply stemming to terms which don't look like
// natural language words.
if (!Db::isSpellingCandidate(*it)) {
LOGDEB1(("createExpansionDbs: skipped: [%s]\n", (*it).c_str()));
continue;
}
// Create stemming synonym for every language. The input is the // Create stemming synonym for every language. The input is the
// lowercase accented term // lowercase accented term
for (unsigned int i = 0; i < langs.size(); i++) { for (unsigned int i = 0; i < langs.size(); i++) {