It looks like the CJK one actually does 2-grams plus a little processing separate processing on latin text.
That's kind of interesting - in general can I build a custom tokenizer from existing tokenizers that treats different parts of the input differently based on the utf-8 range of the characters? E.g. use a porter stemmer for stretches of Latin text and n-gram or something else for CJK? -Peter On Tue, Nov 10, 2009 at 9:21 PM, Otis Gospodnetic <otis_gospodne...@yahoo.com> wrote: > Yes, that's the n-gram one. I believe the existing CJK one in Lucene is > really just an n-gram tokenizer, so no different than the normal n-gram > tokenizer. > > Otis > -- > Sematext is hiring -- http://sematext.com/about/jobs.html?mls > Lucene, Solr, Nutch, Katta, Hadoop, HBase, UIMA, NLP, NER, IR > > > > ----- Original Message ---- >> From: Peter Wolanin <peter.wola...@acquia.com> >> To: solr-user@lucene.apache.org >> Sent: Tue, November 10, 2009 7:34:37 PM >> Subject: Re: any docs on solr.EdgeNGramFilterFactory? >> >> So, this is the normal N-gram one? NGramTokenizerFactory >> >> Digging deeper - there are actualy CJK and Chinese tokenizers in the >> Solr codebase: >> >> http://lucene.apache.org/solr/api/org/apache/solr/analysis/CJKTokenizerFactory.html >> http://lucene.apache.org/solr/api/org/apache/solr/analysis/ChineseTokenizerFactory.html >> >> The CJK one uses the lucene CJKTokenizer >> http://lucene.apache.org/java/2_9_1/api/contrib-analyzers/org/apache/lucene/analysis/cjk/CJKTokenizer.html >> >> and there seems to be another one even that no one has wrapped into Solr: >> http://lucene.apache.org/java/2_9_1/api/contrib-smartcn/org/apache/lucene/analysis/cn/smart/package-summary.html >> >> So seems like the existing options are a little better than I thought, >> though it would be nice to have some docs on properly configuring >> these. >> >> -Peter >> >> On Tue, Nov 10, 2009 at 6:05 PM, Otis Gospodnetic >> wrote: >> > Peter, >> > >> > For CJK and n-grams, I think you don't want the *Edge* n-grams, but just >> n-grams. >> > Before you take the n-gram route, you may want to look at the smart Chinese >> analyzer in Lucene contrib (I think it works only for Simplified Chinese) and >> Sen (on java.net). I also spotted a Korean analyzer in the wild a few months >> back. >> > >> > Otis >> > -- >> > Sematext is hiring -- http://sematext.com/about/jobs.html?mls >> > Lucene, Solr, Nutch, Katta, Hadoop, HBase, UIMA, NLP, NER, IR >> > >> > >> > >> > ----- Original Message ---- >> >> From: Peter Wolanin >> >> To: solr-user@lucene.apache.org >> >> Sent: Tue, November 10, 2009 4:06:52 PM >> >> Subject: any docs on solr.EdgeNGramFilterFactory? >> >> >> >> This fairly recent blog post: >> >> >> >> >> http://www.lucidimagination.com/blog/2009/09/08/auto-suggest-from-popular-queries-using-edgengrams/ >> >> >> >> describes the use of the solr.EdgeNGramFilterFactory as the tokenizer >> >> for the index. I don't see any mention of that tokenizer on the Solr >> >> wiki - is it just waiting to be added, or is there any other >> >> documentation in addition to the blog post? In particular, there was >> >> a thread last year about using an N-gram tokenizer to enable >> >> reasonable (if not ideal) searching of CJK text, so I'd be curious to >> >> know how people are configuring their schema (with this tokenizer?) >> >> for that use case. >> >> >> >> Thanks, >> >> >> >> Peter >> >> >> >> -- >> >> Peter M. Wolanin, Ph.D. >> >> Momentum Specialist, Acquia. Inc. >> >> peter.wola...@acquia.com >> > >> > >> >> >> >> -- >> Peter M. Wolanin, Ph.D. >> Momentum Specialist, Acquia. Inc. >> peter.wola...@acquia.com > > -- Peter M. Wolanin, Ph.D. Momentum Specialist, Acquia. Inc. peter.wola...@acquia.com
