I am not sure what you are after. Google already has a nepali search engine per .np domain. I am pretty sure their spiders are optimized for Unicode, ASCII, latin1, etc.- related character sets. A lot is involved in building a search engine- from crawling, indexing to retrieving; I don't know what you mean by 'tokenize'. If you are a developer and want to learn/improve the way searches work you could write a sample one using 'java.net' package. Or try using 'wget' in UNIX. Moreover, if you want to mimick enterprise-scale search engine I suggest you try Nutch- a free search engine entirely written in Java from Apache: http://lucene.apache.org/nutch/. -JavaBeans