Here are the steps to tune Koha to search Indian languages precisely.
I tried these steps on Koha 3.12.x installation based on Debian packages.
I borrowed these steps from Koustubha Kale posting (http://tinyurl.com/kza7l84) in Koha Discussion forum.
I made slight changes in the instructions. Lets start tuning...

Open a terminal and apply following commands one  by one.
Install additional packages

Install additional packages

sudo su

apt-get install libicu-dev yaz-icu

Copy few files to a new location

cd  /usr/share/idzebra-2.0/tab/

cp icu.idx phrases-icu.xml string.chr words-icu.xml /etc/koha/zebradb/etc/

cp /etc/koha/zebradb/etc/words-icu.xml /etc/koha/zebradb/etc/icu.xml

Open following file,

leafpad /etc/koha/zebradb/etc/icu.xml

change <icu_chain locale="en"> to <icu_chain locale="en_IN.UTF-8">

Folder ownership

sudo chown -R library-koha:library-koha /etc/koha/zebradb/etc

gedit /etc/koha/sites/library/zebra-biblios.cfg

add line at the bottom of the file,

index: icu.idx

Restart zebra server

sudo su 

koha-restart-zebra library

Rebuild zebra index

sudo koha-rebuild-zebra -f -v library

Turn Off

Home → Administration > System Preferences > Searching > QueryFuzzy