<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
From: <b class="gmail_sendername">L. Amber Wilcox-O'Hearn</b> <span
dir="ltr"><<a moz-do-not-send="true"
href="mailto:amber.wilcox.ohearn@gmail.com">amber.wilcox.ohearn@gmail.com</a>></span><br>
<blockquote
cite="mid:CALFkT+=rPHjZj6K4aRdLgbsTEhrOX3ZOw+TLdXNXjc+-U4vqAA@mail.gmail.com"
type="cite">
<div class="gmail_quote"><br>
Thank you, Andreas. I wasn't aware of these capabilities.<br>
<br>
The server-port worked exactly as expected. That is, if I give
it w1<br>
w2 w3, it returns p(w3|w1w2). Combined with the caching, it
looks<br>
very promising for my applications.<br>
<br>
The other solution using -counts (or actually -ppl for my case)
also<br>
worked, but of course if I give it w1 w2 w3, it returns the<br>
probability of that whole string, i.e. p(w1) * p(w2|w1) *
p(w3|w1w2),<br>
which would be redundant for my purposes.<br>
</div>
</blockquote>
That's not correct. ngram -counts will output CONDITIONAL ngram
probabilities.<br>
<dl>
<dt><b>-counts</b><i> countsfile</i>
</dt>
<dd>
Perform a computation similar to <b>-ppl</b>,
but based only on the N-gram counts found in <i>countsfile</i>.
Probabilities are computed for the last word of each N-gram,
using the
other words as contexts, and scaling by the associated N-gram
count.
Summary statistics are output at the end, as well as before each
escaped input line.
</dd>
</dl>
So it should do exactly what you need.<br>
<br>
Andreas<br>
<br>
</body>
</html>