# Lingke: A Fine-grained Multi-turn Chatbot for Customer Service

**Authors:** Pengfei Zhu, Zhuosheng Zhang, Jiangtong Li, Yafang Huang, Hai Zhao  
**arXiv:** [1808.03430](https://arxiv.org/abs/1808.03430)  
**Date:** 2018-08-10  
**Summary page:** https://www.summarizepaper.com/en/paper/1808.03430v1/

## Key points

- Lingke is a chatbot designed to address the limitations of traditional chatbots in multi-turn conversations
- Traditional chatbots struggle in multi-turn scenarios and rely on human dialogue data
- Lingke answers questions based on product introduction documents and engages in multi-turn conversations
- Lingke uses a fine-grained pipeline processing approach to distill responses from unstructured documents
- Attentive sequential context-response matching techniques are introduced to enhance Lingke's ability in handling multi-turn conversations
- Lingke overcomes challenges faced by traditional chatbots and improves performance in multi-turn scenarios
- Information retrieval techniques and attentive context-response matching are leveraged by Lingke for accurate and relevant responses during customer service interactions
- Lingke combines information retrieval and attentive matching mechanisms to enhance its performance in addressing customer queries and engaging in extended conversations

## Summary

The paper titled "Lingke: A Fine-grained Multi-turn Chatbot for Customer Service" introduces Lingke, an information retrieval augmented chatbot designed to address the limitations of traditional chatbots in handling multi-turn conversations. While traditional chatbots rely on a large amount of human dialogue data, they often struggle to provide satisfactory performance in multi-turn scenarios. Lingke is specifically developed to answer questions based on product introduction documents and effectively engage in multi-turn conversations. The authors propose a fine-grained pipeline processing approach that distills responses from unstructured documents. Additionally, they introduce attentive sequential context-response matching techniques to enhance the chatbot's ability to handle multi-turn conversations. The authors highlight that Lingke's approach overcomes the challenges faced by traditional chatbots and improves their performance in multi-turn scenarios. By leveraging information retrieval techniques and attentive context-response matching, Lingke demonstrates its capability to provide more accurate and relevant responses during customer service interactions. Overall, this paper presents Lingke as an advanced chatbot solution that combines information retrieval and attentive matching mechanisms to enhance its performance in addressing customer queries and engaging in extended conversations. This makes it well suited for customer service applications where accuracy and relevance are critical factors for success.