I'm trying to run the logon parse ranking experiments on a corpus smaller than the jhpstg.<br><br>I want to create a virtual profile that contains just jh0 through jh3. I created a directory tsdb/home/jh03 that contains the file virtual with contents:<br>
<br><div style="margin-left: 40px;">"jh0"<br>"jh1"<br>"jh2"<br>"jh3"<br></div><br>I then created a file called fc.jh03.lisp with the following contents:<br><br><div style="margin-left: 40px;">
(in-package :tsdb)<br><br>(load "parsing.lisp")<br><br>(operate-on-profiles '("jh03"))<br><br></div>I ran load --binary fc.jh03.lisp. This process ran for about 3 hours. When it was done, it had created an fc.mlm file in the tsdb/home/jh03 directory as I would have expected. However it also created various <br>
<br><div style="margin-left: 40px;">[jh03] GP[4] +PT -LEX CW[2] +AE NS[4] NT[type] +NB LM[0] FT[:::1] RS[] MM[tao_lmvm] ...<br></div><br>directories with varying values for RT, AT, and VA. These directories contain empty fold and score files. I was surprised by the creation of these directories, since I expect them to be created by the grid step and not the feature caching step.<br>
<br>Is this a bug?<br clear="all"><br>-- <br>Bill McNeill<br><a href="http://staff.washington.edu/billmcn/index.shtml">http://staff.washington.edu/billmcn/index.shtml</a><br>Sent from: Seattle Washington United States.