![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Fredkin speaking at CMU
Mar. 5th, 2007 12:26 pmThis wednesday at 12:30pm, Ed Fredkin is going to speak on NTS: A New Transportation System, at NSH 1305. Ed is an amazing guy. I met him in 2003 (See this entry. Earlier that day, I met these people)
Mar. 5th, 2007
argumentative structures project
Mar. 5th, 2007 06:42 pmMy argumentative structures project has been kinda stuck. I can't think of anything interesting to do with the data I have.
Today I spoke to Wilfried Sieg, who said that the project of doing machine learning on inferences in logic textbooks (tagged with inference type) would be interesting to him, even if it isn't interesting to William. For one thing, this could provide automatic grading for logic courses.
Journals that would be interested in such work:
* Teaching Philosophy
* NA-CAP (this year in Chicago, proposals due March 21)
I don't find this task terribly interesting, but maybe it could open the ground for more interesting things, as we slowly make progress towards modeling arguments in real-world texts. I should look for prior work. But it's still unclear how learning on logic textbooks could transfer enough to get us reasonable performance on real-world texts. Do we even have intermediate corpora between formalistic logic textbooks and real-world argumentative texts, with which to guide this learning?
Today I spoke to Wilfried Sieg, who said that the project of doing machine learning on inferences in logic textbooks (tagged with inference type) would be interesting to him, even if it isn't interesting to William. For one thing, this could provide automatic grading for logic courses.
Journals that would be interested in such work:
* Teaching Philosophy
* NA-CAP (this year in Chicago, proposals due March 21)
I don't find this task terribly interesting, but maybe it could open the ground for more interesting things, as we slowly make progress towards modeling arguments in real-world texts. I should look for prior work. But it's still unclear how learning on logic textbooks could transfer enough to get us reasonable performance on real-world texts. Do we even have intermediate corpora between formalistic logic textbooks and real-world argumentative texts, with which to guide this learning?