Settings for the index string extraction¶
This page is generated by Machine Translation from Japanese.
About the index string extraction¶
You must isolate the document in order to register as the index when creating indexes for the search. Tokenizer is used for this.
Basically, carved by the tokenizer units smaller than go find no hits. For example, statements of living in Tokyo, Japan. Was split by the tokenizer now, this statement is in Tokyo, living and so on. In this case, in Tokyo, Word search, you will get hit. However, when performing a search with the word ‘Kyoto’ will not be hit. For selection of the tokenizer is important.
You can change the tokenizer by setting the schema.xml analyzer part is if the Fess in the default CJKTokenizer used.
CJKTokenizer index bi-gram, in other words two characters to like Japan Japanese multibyte string creates. In this case, can’t find one letter words.
StandardTokenizer creates index uni-gram, in other words one by one for the Japan language of multibyte-character strings. Therefore, the less search leakage. Also, with StandardTokenizer can’t CJKTokenizer the search query letter to search to. However, please note that the index size increases.
The following example to change the analyzer part like solr/core1/conf/schema.xml, you can use the StandardTokenizer.
<schema name="fess" version="1.4"> <types> <fieldType name="text" class="solr.TextField" positionIncrementGap="100"> <analyzer type="index"> <charFilter class="solr.MappingCharFilterFactory" mapping="mapping_ja.txt"/> <tokenizer class="solr.StandardTokenizerFactory"/> : </analyzer> <analyzer type="query"> <charFilter class="solr.MappingCharFilterFactory" mapping="mapping_ja.txt"/> <tokenizer class="solr.StandardTokenizerFactory"/> :
Also, useBigram is enabled by default in the webapps/fess/WEB-INF/classes/app.dicon change to false.
: <component name="queryHelper" class="jp.sf.fess.helper.impl.QueryHelperImpl"> <property name="useBigram">true</property> :
After the restart the Fess .