TY  - GEN
A1  - Galke, Lukas
A1  - Gerstenkorn, Gunnar
A1  - Scherp, Ansgar
T1  - A case atudy of closed-domain response suggestion with limited training data
T2  - Database and Expert Systems Applications : DEXA 2018 Iinternational workshops
N2  - We analyze the problem of response suggestion in a closed domain along a real-world scenario of a digital library. We present a text-processing pipeline to generate question-answer pairs from chat transcripts. On this limited amount of training data, we compare retrieval-based, conditioned-generation, and dedicated representation learning approaches for response suggestion. Our results show that retrieval-based methods that strive to find similar, known contexts are preferable over parametric approaches from the conditioned-generation family, when the training data is limited. We, however, identify a specific representation learning approach that is competitive to the retrieval-based approaches despite the training data limitation.
Y1  - 2018
UR  - https://publishup.uni-potsdam.de/frontdoor/index/index/docId/54072
SN  - 978-3-319-99133-7
SN  - 978-3-319-99132-0
SN  - 1865-0929
SN  - 1865-0937
VL  - 903
SP  - 218
EP  - 229
PB  - Springer
CY  - Berlin
ER  -