[Corpora-List] 2nd CFP: 4th Workshop on Argument Mining at EMNLP 2017

Habernal, Ivan habernal at ukp.informatik.tu-darmstadt.de
Tue Apr 18 16:04:41 CEST 2017

Distinguishing between "what" (factoid) and "why" (non-factoid) question from the perspective of information solicitation has been an active topic in Question Answering (https://web.stanford.edu/class/cs124/lec/qa.pdf). Argument mining is not much about giving the right answer to my decision-making question but rather finding arguments with their claims and how they are supported. So there might be some overlap but I think these two tasks are different...



> Am 18.04.2017 um 14:56 schrieb John F Sowa:
>> On 4/18/2017 5:03 AM, Habernal, Ivan wrote:
>>> Argument mining ... involves automatically identifying
>>> argumentative structures within discourse...
>> One point I have not seen in the literature about arguments is
>> the distinction between "What?" questions and "Why?" questions.
>> A what-question asks for information. But a why-question asks
>> for a reason or sequence of reasons that lead to a conclusion
>> or an explanation.
>> Terry Winograd's SHRDLU system from the early '70s could answer
>> why-questions by tracing backwards through the its theorem prover.
>> But documents written by and for humans (other than mathematicians
>> or medieval Scholastics) are rarely derived by a theorem prover.
>> Has anyone been addressing the what/why distinction?
>> John

More information about the Corpora mailing list