After hearing complaints that the LSJ lexicon had a low hit ratio I started testing on texts other than my own and found the complaints to have merit. This is because Greek texts are not consistent when it comes to how unicode for polytonic Greek is applied. A unicode text can be encoded using composing diacritics, or precomposed diacritics. The consensus today favors precomposed diacritics but within this subset, due to faulty duplication, we again have two choices. Vowels can be represented using acute or tonos with the consensus being in favor of the tonos. Most computer operating can disambiguate these variations but devices like the Kindle and Android based systems do not.
I have re-encoded all of my texts including the LSJ to follow this consensus.
Search of the LSJ from Xenophon Anabasis yields a 95% hit ratio excluding words with two accents (these must be typed in using beta code transliteration).
I would be happy to help anyone re encode texts to get them to work with the LSJ.
I have also updated the LSJ reducing its size to 50MB (Amazon limit). To do this I had to remove the grammatical parsing information within the inflection section that had been added to the lexicon. At the same time I have substantially increased the inflection database.
For those confused when search on inflected forms brings up only the headword, please continue by searching on the headword. This was the the only solution to the problem of multiple headwords branching from an single inflection and had to be applied universally.
http://www.amazon.com/Complete-Liddell-Lexicon-Inflections-ebook/dp/B0092K07BQPlease note that I have oversimplified the extremely important topic of text encoding and expect "disagreement" about what constitutes a "consensus". Below are a few websites on the subject and I strongly urge all become familiar with this important topic.
http://www.tlg.uci.edu/~opoudjis/unicode/unicode.htmlhttp://www.stoa.org/unicode/http://unicode.org/faq/normalization.html