id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
cord-026960-g844u7xg	Wang, Disen	An Adaptive Response Matching Network for Ranking Multi-turn Chatbot Responses	2020-05-26		.txt	text/plain	4523	237	52	To address this limitation, this paper proposes an adaptive response matching network (ARM) to better model the matching relationship in multi-turn conversations. Specifically, the Dual-encoder model [8] used two LSTMs to generate the embeddings for the utterances and candidate response respectively to compute the matching score. Deep attention model (DAM) [20] proposed self-attention and cross-attention to construct semantic representations at different granularity, and the multirepresentation fusion network (MRFN) [13] has further applied multiple representation strategies for utterances and fused them in the final step to compute the matching scores. Few of them studied how to adapt the matching model to different types of utterances and how to incorporate the domain knowledge in a more general way, which is the focus of our paper. Although a few models have been proposed to solve the problem of multi-turn response selection [2, 13, 17, 20] , none of them studied how to directly adapt the matching mechanisms to different utterance types.	./cache/cord-026960-g844u7xg.txt	./txt/cord-026960-g844u7xg.txt