# Towards an Engine for Lifelong Interactive Knowledge Learning in Human-Machine Conversations

**Authors:** Sahisnu Mazumder, Nianzu Ma, Bing Liu  
**arXiv:** [1802.06024](https://arxiv.org/abs/1802.06024)  
**Date:** 2018-02-16  
**Summary page:** https://www.summarizepaper.com/en/paper/1802.06024v1/

## Key points

- Chatbots have gained popularity for automating conversations
- Limitations of chatbots include their inability to learn new knowledge during conversations
- Authors propose building a general knowledge learning engine for chatbots
- Engine enables chatbots to continuously learn during conversations
- Technique called lifelong interactive learning and inference (LiLi) is introduced
- LiLi imitates how humans acquire knowledge and perform inference during conversations
- Experimental results demonstrate the promise of LiLi in enhancing chatbot capabilities

## Summary

In recent years, chatbots have gained popularity due to their ability to automate conversations. However, they still have limitations that restrict their applications; one major weakness being their inability to learn new knowledge during conversations. Their knowledge is fixed beforehand and cannot be expanded or updated in real-time. To address this issue, the authors propose building a general knowledge learning engine for chatbots that enables them to continuously learn during conversations. Over time, the chatbots become more knowledgeable and proficient in learning and conversation. The authors model this task as an open-world knowledge base completion problem and introduce a novel technique called lifelong interactive learning and inference (LiLi) to solve it. LiLi imitates how humans acquire knowledge and perform inference during interactive conversations. Experimental results demonstrate the high promise of LiLi in enhancing chatbot capabilities. The paper titled "Towards an Engine for Lifelong Interactive Knowledge Learning in Human-Machine Conversations" by Sahisnu Mazumder, Nianzu Ma, and Bing Liu provides further details on this approach.