<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<div class="moz-cite-prefix">On 5/28/2013 12:42 AM, 贺天行 wrote:<br>
</div>
<blockquote
cite="mid:CANyb1j3uoGG00tGu-t00bFhZ_BKWyqA1_E74V99VJgPs2KkQeg@mail.gmail.com"
type="cite">
<div dir="ltr">Hello
<div style="">When I use order 2 kndiscount, I get a unigram
model and a bigram model</div>
<div style="">Then I use order 1 kndiscount, I also get a
unigram</div>
<div style="">But these two unigrams are different, I read the </div>
<div style=""><a moz-do-not-send="true"
href="http://www.speech.sri.com/projects/srilm/manpages/ngram-discount.7.html">http://www.speech.sri.com/projects/srilm/manpages/ngram-discount.7.html</a><br>
</div>
<div style="">It seems that this has to do with some
implementation issue, what i want to ask is, is the unigram I
get in the order 2 kndiscount uncorrect?</div>
<div style=""><br>
</div>
<div style="">Because if I use order1 Katz discount and order 2
Katz discount, the two unigrams are the same, so I think I
need to treat kndiscount result with caution.</div>
</div>
</blockquote>
<br>
It is one of the distinguishing features of KN discounting that the
lower-order (backoff) distributions are estimated differently from
the highest-order distribution.<br>
You are not supposed to use the unigram distribution in a
KN-smoothed bigram by itself. <br>
<br>
So what you're seeing is completely expected and correct. For a
detailed explanation see the <a
href="http://www.speech.sri.com/projects/srilm/manpages/pdfs/chen-goodman-tr-10-98.pdf">Chen
and Goodman paper.</a><br>
<br>
Andreas<br>
<br>
</body>
</html>