If you want to define your own custom settings for SOLR analyzers and filters, then modify the installation script ( src/server/manageSchema), which calls the SOLR API, accordingly.įor more information, consult the SOLR documentation, e.g. Therefore, applying ascii folding while indexing doesn’t prevent us from having results with accents. The analyzer is not related to what is stored in the collection but how this collection is indexed, as indexes are stored aside from the collection. Note: If the analyzer has no type, then it is applied to both index & query. The SOLR filters used in SFS by default are the following: Note also that different scoring approaches can be applied to the synonym terms depending on their occurence. Dutch: solr.SnowballPorterFilterFactoryįor more information, consult the SOLR documentation, e.g.Italian: solr.ItalianLightStemFilterFactory.French: solr.FrenchLightStemFilterFactory.German: solr.GermanLightStemFilterFactory.The SOLR synonym filters used by default in SFS are: Remember that synonyms need to be defined per language. If you want to develop your own custom approach for pushing synonyms to SOLR, e.g., directly from a database, then modify the installation script ( src/server/manageSchema), which calls the SOLR API, accordingly. Synonyms are automatically ASCII-folded by the script (see below). a the mages of the United Kingdom query would be expanded to: the (mages OR sorcerer OR sorceress) of the (United Kingdon OR Great Britain OR England). Synonyms are applied on the search query terms, e.g. For more information, see Manage synonyms. For more information, consult the SOLR documentation. Remember that lemmatisation need to be defined per language. In order to customise lemmatisation filters, modify the installation script ( src/server/manageSchema), which calls the SOLR API, accordingly. So search results that contain the query terms as an exact phrase (all the terms and in the exact sequential order) will get an additional score boost.ĭoc2 = field type provided by Solr. It gives you the possibility of a multiplicative factor (the weight) to the score calculated by Solr for the query match in a field.īoosting phrase matches, so documents containing phrase matches of the user query will be boosted in accordance with the factor specified This boosting is equivalent to the one currently in place. For more information, see: Relevance of free-text search results: How to boost specific dataflowsĪpart from the boost queries currently in use, moving to edismax query parsing would give additional possibilities in terms of:īoosting fields (so the weight passed as a parameter will end up as a qf Solr parameter) Specific documents (in our case dataflows) can be boosted in the scoring algorithm. For more information, see: Relevance of free-text search results: How to tweak the weights of specific dataflow properties Specific document fields (in our case dataflow properties) can be boosted in the scoring algorithm. Starting from Solr 6, Lucene implements by default the probabilistic IR model Okapi BM25, a scoring function that make use of the features forementioned.įor more information on Lucene/Solr scoring:
0 Comments
Leave a Reply. |