<p>Why not write a conversion adapter from shift-JIS to Unicode and put that conversion in your data pipeline after chasen? (possibly with a unicode-to-shift-JIS adapter upstream as well?)</p>
<p>Much of NLP work is plumbing; this is even a pretty easy piece of plumbing: very little chewing-gum required!</p>
<p>Jeremy<br>
</p>
<div class="gmail_quote">On Jun 28, 2011 9:50 AM, "Laurence Anthony" <<a href="mailto:anthony0122@gmail.com">anthony0122@gmail.com</a>> wrote:<br type="attribution">>><br>>> A Japanese user of WordSmith needs help with the Chasen software, which I<br>
> understand provides segmentation of the string of characters in Japanese.<br>> Desired output form would be UTF16 for WordSmith.<br>>><br>>> Can anyone advise, please? Is this possible?<br>>><br>
>> Mike<br>> <br>> <br>> Hi Mike,<br>> <br>> I think Chasen only outputs to ANSI (SHIFT-JIS here in Japan) or UTF-8.<br>> However, an alternative tool is MeCab, which does offer tentative UTF-16<br>
> support.<br>> <br>> You can read about it here (unfortunately everything is in Japanese):<br>> <a href="http://mecab.sourceforge.net">http://mecab.sourceforge.net</a><br>> <br>> Here's a summary of the latest version (dated 2009):<br>
> 2009-09-27 MeCab 0.98<br>> UTF16$B$N%5%]!<%H(B($B<B83E*(B)<br>> Windows$BHG$G$NJ8;z%3!<%IJQ49$K(B MutlByteToWideChar$BEy$N(B Native API$B$r;H$&$h$&$KJQ99(B<br>> $B%=!<%9%3!<%I$r(B Google coding style $B$KJQ99(B<br>> $B%U%)!<%^%C%H;XDj$G(B EON (end of N-best) $B$NDI2C(B (-S or --eon-format)<br>
> Shift-JIS$B4D6-$GH>3Q%+%?%+%J$N07$$$KLdBj$,$"$C$?$N$r=$@5(B<br>> online learning $B$N%5%]!<%H(B ($B<B83E*(B)<br>> Wno-deprecated$B$r$D$1$J$/$F$b%3%s%Q%$%k$G$-$k$h$&$K$7$?(B<br>> $B:Y$+$$%P%0$N=$@5(B<br>> <br>> Hope that helps!<br>> Laurence.<br></div>