<div dir="ltr"><div><div><div>Dear all,<br><br></div>Here is a further mystery from the annals of the MT lab in UW's 567 --- there's a work-around that's satisfactory, but Woodley and I were not able to get to the bottom of the case, so I thought I'd post the details in case anyone is intrigued enough to investigate.<br> </div><div><br></div>The attached frr.tgz and eng.tgz grammars are both set up for the lightweight LKB-to-LKB translation set up that the students use in my grammar engineering class to work out their transfer rules. This set up is admittedly odd in that it embeds the transfer grammars in the source side monolingual grammar, meaning that "x" et al have to be define in such a way that they work for both.<br> <br>The mystery we're facing is that in the frr grammar (but not the very similar eng grammar), a transfer rule that identifies a variable as 'x' type won't fire. If the variable is only an 'i', it works. Using the "Debug" menu on the transfer output windows, we were able to see that the feature SCRATCH, DITCH et al show up on the xes in the transfer rules for frr but not for eng. I suspect this is related to the problem, but we didn't have time to work out why the grammars differ in this way (and thus be able to test that hypothesis).<br> <br></div>This behavior can (I hope!) be reproduced by parsing "ik driuwe di" with the frr grammar in the LKB and then selecting "Rephrase". The pro-drop rule does not apply (in contrast to "I chase you" in the eng grammar). If the pro-drop rule is edited to not identify the ARG0 as being of type x, it does apply.<br> <br>Emily<br><div><div><div><br><div>-- <br>Emily M. Bender<br>Associate Professor<br>Department of Linguistics<br>Check out CLMS on facebook! <a href="http://www.facebook.com/uwclma" target="_blank">http://www.facebook.com/uwclma</a><br> </div></div></div></div></div>