Modeling Multi-turn Conversation with Deep Utterance Aggregation

Multi-turn conversation understanding is a major challenge for building intelligent dialogue systems. This work focuses on retrieval-based response matching for multi-turn conversation whose related work simply concatenates the conversation utterances, ignoring the interactions among previous utterances for context modeling. In this paper, we formulate previous utterances into context using a proposed deep utterance aggregation model to form a fine-grained context representation. In detail, a self-matching attention is first introduced to route the vital information in each utterance. Then the model matches a response with each refined utterance and the final matching score is obtained after attentive turns aggregation. Experimental results show our model outperforms the state-of-the-art methods on three multi-turn conversation benchmarks, including a newly introduced e-commerce dialogue corpus.

PDF Abstract COLING 2018 PDF COLING 2018 Abstract

Datasets


Introduced in the Paper:

E-commerce

Used in the Paper:

Douban UDC Douban Conversation Corpus
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Conversational Response Selection Douban DUA MAP 0.551 # 14
MRR 0.599 # 15
P@1 0.421 # 15
R10@1 0.243 # 15
R10@2 0.421 # 14
R10@5 0.780 # 14
Conversational Response Selection E-commerce DUA R10@1 0.501 # 11
R10@2 0.700 # 11
R10@5 0.921 # 11
Conversational Response Selection Ubuntu Dialogue (v1, Ranking) DUA R10@1 0.752 # 19
R10@2 0.868 # 18
R10@5 0.962 # 18

Methods


No methods listed for this paper. Add relevant methods here