HI Vanessa,
You’re correct. Dspace uses Apache Lucene to index and analyse full text, and one of the things a Lucene analyser (note: all docs and code refer to the American spelling, analyzer) does is perform “stemming” on indexed tokens, so common suffixes like “ing”, “ed”, “es”, “ly” are chopped off the tokens, and off your search terms as well. It’s possible to write your own Lucene analyser or extend your own, and replace the default DSAnalyzer that Dspace uses with your version. I’ve done this in my installations, not for stemming, but to properly tokenise macronised vowels (ā ē ī ō ū) that are used in New Zealand but aren’t in supported ISO character sets. This page might help explain the concepts better than myself: http://wiki.apache.org/lucene-java/ConceptsAndDefinitions?highlight=(stemmer) This is a quote from the Dspace system docs: The Lucene analyzer used in searching and indexing can be configured by setting the search.analyzer configuration item in dspace.cfg to the class of the desired analzyer. If this item is not present/commented out, the default Lucene analyzer org.dspace.search.DSAnalyzer is used. As well as those analyzers included in the Lucene distribution (see lucene.jar), a Chinese analyzer from the Lucene sandbox is included in lucene-sandbox.jar. This analyzer is yet to be included in the core Lucene distribution but can be configured by setting search.analyzer = org.apache.lucene.analysis.cn.ChineseAnalyzer in dspace.cfg. Cheers, Kim From: dspace-general-boun...@mit.edu [mailto:dspace-general-boun...@mit.edu] On Behalf Of Vanessa Barrett Sent: Friday, 18 September 2009 4:30 p.m. To: dspace-general@mit.edu Subject: [Dspace-general] Indexing - automatic mapping of plurals andalternate endings? Can anyone confirm my understanding of how DSpace performs keyword indexing/searching? I suspect that it is doing automatic mapping of singular and plural forms of words. How I came to this understanding was as follows. I was searching for an item authored by Alys (alternate spelling of Alice) Clark. I retrieved three items none of which had the word alys in the metadata or bitstream. If I searched for alys on its own I got 168 hits and a cursory glance at the results list showed that they all had an author with some part of their name being ali. I then tried searching for each of the following forms aly, ali, alis, alys alies All of these as single search terms retrieved exactly the same number of records – 168. Results included items with the following strings in Abstract - ALIS (Advanced Landmine Imaging System), which is a novel landmine detection sensor system - Current ventilatory practices for the management of ALI favor low tidal volumes - Current Trends in Periodontal Diagnosis & Disease Recognition in Malaysia / T.B. Taiyeb Ali - Radiology in the acute abdomen / P.G. Devitt, A. Aly, M. Thomas My conclusion is that DSpace is doing some process of mapping plural to singular forms of words including allowing for alternate endings. If it is doing this it is very clever but just a little annoying as Alys is not the plural of Ali. Also if clever enough to do this why can’t it map fiber to fibre and color to colour which would have much greater benefits in searching a database that includes North American and European data. Cheers, Vanessa Barrett Digital Services Librarian The University of Adelaide, AUSTRALIA 5005 Ph : +61 8 8303 4625 e-mail: vanessa.barr...@adelaide.edu.au CRICOS Provider Number 00123M ----------------------------------------------------------- IMPORTANT: This message may contain confidential or legally privileged information. If you think it was sent to you by mistake, please delete all copies and advise the sender. For the purposes of the SPAM Act 2003, this email is authorised by The University of Adelaide. Think green: read on the screen.
_______________________________________________ Dspace-general mailing list Dspace-general@mit.edu http://mailman.mit.edu/mailman/listinfo/dspace-general
