Abstract:Building an intelligent dialogue system with the ability to select a proper response according to a multi-turn context is challenging in three aspects: (1) the meaning of a context–response pair is built upon language units from multiple granularities (e.g., words, phrases, and sub-sentences, etc.); (2) local (e.g., a small window around a word) and long-range (e.g., words across the context and the response) dependencies may exist in dialogue data; and (3) the relationship between the context and the response candidate lies in multiple relevant semantic clues or relatively implicit semantic clues in some real cases. However, existing approaches usually encode the dialogue with mono-type representation and the interaction processes between the context and the response candidate are executed in a rather shallow manner, which may lead to an inadequate understanding of dialogue content and hinder the recognition of the semantic relevance between the context and response. To tackle these challenges, we propose a representation [ K ] -interaction [ L ] -matching framework that explores multiple types of deep interactive representations to build context-response matching models for response selection. Particularly, we construct different types of representations for utterance–response pairs and deepen them via alternate encoding and interaction. By this means, the model can handle the relation of neighboring elements, phrasal pattern, and long-range dependencies during the representation and make a more accurate prediction through multiple layers of interactions between the context–response pair. Experiment results on three public benchmarks indicate that the proposed model significantly outperforms previous conventional context-response matching models and achieve slightly better results than the BERT model for multi-turn response selection in retrieval-based dialogue systems.

Keep and Select: Improving Hierarchical Context Modeling for Multi-Turn Response Generation

How to Represent Context Better? an Empirical Study on Context Modeling for Multi-turn Response Selection.

Improving Matching Models with Hierarchical Contextualized Representations for Multi-turn Response Selection

Deep context modeling for multi-turn response selection in dialogue systems

Learning to Detect Relevant Contexts and Knowledge for Response Selection in Retrieval-based Dialogue Systems

Hierarchical matching network for multi-turn response selection in retrieval-based chatbots

Learning to Select the Relevant History Turns in Conversational Question Answering

CoHS-CQG: Context and History Selection for Conversational Question Generation

X-RECOSA: Multi-Scale Context Aggregation for Multi-Turn Dialogue Generation

Pay More Attention to History: A Context Modeling Strategy for Conversational Text-to-SQL.

Pay More Attention to History: A Context Modelling Strategy for Conversational Text-to-SQL

Response Ranking with Multi-types of Deep Interactive Representations in Retrieval-based Dialogues

Query Enhanced Knowledge-Intensive Conversation via Unsupervised Joint Modeling

Topic-Aware Multi-turn Dialogue Modeling

Context-Controlled Topic-Aware Neural Response Generation for Open-Domain Dialog Systems

HKA: A Hierarchical Knowledge Attention Mechanism for Multi-Turn Dialogue System

Contextual Fine-to-Coarse Distillation for Coarse-grained Response Selection in Open-Domain Conversations

Evaluating and Improving Context Attention Distribution on Multi-Turn Response Generation using Self-Contained Distractions

A Sequential Matching Framework for Multi-Turn Response Selection in Retrieval-Based Chatbots

Learning Multi-turn Response Selection in Grounded Dialogues with Reinforced Knowledge and Context Distillation

CoE-SQL: In-Context Learning for Multi-Turn Text-to-SQL with Chain-of-Editions