Machine Learning Based Question Answering

 

Seminar
Computational Lingustics
Master Programme
Summer Semester 2016


 

News

Don't Worry Be Happy!
 

General Information

Moderator: Günter Neumann, Georg Heigold

Abstract:

Machine Learning has become a major tool for building state-of-the-art question answering systems, i.e., systems that are able to learn to answer natural language questions of divere nature, from factoid to non-factoid questions. In this seminar, we will study state of the art methods and technology for question answering.

Seminar Language: English

Available Certificate Modalities:

Placement in Study Programme:

 

Schedule

Session Number
Date
Topic
Speaker
1
27.04.2016
Organisational meeting
Günter Neumann, Georg Heigold
2
04.05.2016
Overview Question Answering
Günter Neumann
3
11.05.2016
Overview Deep Learning
Georg Heigold
4
18.05.2016
Distant Supervision
Fraser Bowen
5
25.05.2016
Open Question Answering I
Gareth Dwyer
6
01.06.2016
Open Question Answering II
Yiling Chung
7
08.06.2016
QA Corpora
Lukas Schmitt
8
15.06.2016
Answer Selection
Eran Raveh
9
22.06.2016
End-to-end QA
Yauhen Klimovich
11
29.06.2016
Modularity
Stephanie Lund
10
06.07.2016
Memory Models for QA
Jakub Hajic
12
13.07.2016
Modularity
Nima Nabizadeh
14
20.07.2016
Multimedia QA
Mahsa Vafayi
13
27.07.2016
Evaluation
Stalin Varanasi

Please click on the session number to jump to the corresponding references. If available, the topics of the presentations will be linked to the slides of the presentations.

 

References

Session 2 & 3: Overview

Session 4: Distant Supervision

Min et al. (2013) Distant Supervision for Relation Extraction with an Incomplete Knowledge Base, NAACL, 2013.

Session 5: Open Question Answering I

Fader et al. (2014) Open Question Answering Over Curated and Extracted Knowledge Bases, KDD, 2014.

Session 6: Open Question Answering II

Bordes et al. (2014) Open Question Answering with Weakly Supervised Embedding Models, ECML PKDD, 2014.

Session 7: QA Corpora

Serban et al. (2016) Generating Factoid Questions With Recurrent Neural Networks: The 30M Factoid Question-Answer Corpus, arXiv, 2016.

Session 8: Answer Selection

Dos Santos et al. (2016) Attentive Pooling Networks, arXiv, 2016.

Session 9: End-to-end QA

Golub and He (2016) Character-Level Question Answering with Attention, arXiv, 2016.

Session 10: Memory Models for QA

Xiong et al. (2016) Dynamic Memory Networks for Visual and Textual Question Answering, arXiv, 2016.

Session 11: Modularity

Andres et al. (2016) Learning to Compose Neural Networks for Question Answering, arXiv, 2016.

Session 12: Modularity

Jurczyk and Choi (2016) Multi-Field Structural Decomposition for Question Answering, arXiv, 2016.

Session 13: Evaluation

Trec (2015) TREC LiveQA 2015, NIST, 2015.

Session 14: Multimedia QA

Mostafazadeh et al. (2016) Generation of Natural Questions About an Image, arXiv, 2016.

 

Written Report

Students enrolled in the Master's programme can choose to submit a written report (see available certificate modalities). The length of the written report is restricted to eight pages, disregarding bibliographical sources. For this purpose, the linked conference-style template should be used (available for Latex and MS Word). The submission deadline is 30st September 2016. The written report should have the the style of conference proceedings. We expect you to digest the material related to your topic and perform further research. In your report, you should add value to the available information by comparing, criticizing, and highlighting plus points. We want to encourage you to think and develop your own opinion, and will disapprove of copy-pasting. If you have questions on the written report, we will be happy to help you.

You can turn in your report in electronic form as PDF file. Electronic copies should be submitted via e-mail to the following addresses: neumann@dfki.de, georg.heigold@dfki.de