NLPIR/ICTCLA2018 ACADEMIC SEMINAR 4th ISSUE COMPLETED

NLPIR/ICTCLA2018 ACADEMIC SEMINAR 4th ISSUE COMPLETED自然语言处理与信息检索共享平台B1A/m3HT

 自然语言处理与信息检索共享平台}1d”`D;e9C jt[

        Last week, Wang Gang gave us a presentation on the topic of Know What You Don’t Know: Unanswerable Questions for SQuAD.

Dp”yWSP D*@0

 自然语言处理与信息检索共享平台C\^4t a!w0C

        After the presentation, several questions were asked. The Q&As are listed as following:

K]7h6~{({}0

        Q: Why ‘rank sentences that answer a question higher than sentences that do not’ ?自然语言处理与信息检索共享平台$_
]Ei
C cf:sm

        A: This is a strategy used while constructing sentence selection datasets. Detailed information can be learned from the following paper.

$^C!w
DV0

        Wang, N. A. Smith, and T. Mitamura. 2007. What is the jeopardy model? a quasi-synchronous grammar for QA. In Empirical Methods in Natural Language Processing (EMNLP).自然语言处理与信息检索共享平台8l/r’@D-EY-@9u-@ } Xc

 

(Kl|-y/F0

        Q: How do you think about this dataset ?自然语言处理与信息检索共享平台z kS$e/`8LFq o+LO

        A: I think it is very good. Compared with SQuAD 1.1, this new version challenge all the current models. The authors hope their work could encourage the development of new reading comprehension models that know what they don’t know, and therefore understand language at a deeper level.自然语言处理与信息检索共享平台6G;K!V”~cpOH$Pl

 

cS tnX”~C0

        Q: Why the score may be lower at SQuAD 2.0 than 1.1?自然语言处理与信息检索共享平台e5MSl*e&Y

        A: There are more unanswerable questions in SQuAD 2.0. So the gap between human and machine is getting bigger.

|Y`3PF-S0

 

Y&Z+|%fK”D*o y0

The powerpoint of the presentation can be download.自然语言处理与信息检索共享平台E/{4Z_mP%n U

 自然语言处理与信息检索共享平台W;o”NH Z

  4_powerpoint.pdf(882 KB)自然语言处理与信息检索共享平台8|Z \:{H1b6~n

IMG_1256IMG_1256

You May Also Like

About the Author: nlpir

发表回复