朱哲人 刘敏 陈鹏

[摘    要]银行业是典型的知识密集领域，各部门因业务规则及系统的持续更新、各层级人员流动等因素产生了对业务答疑的知识复用需求。随着深度学习、机器学习在自然语言处理（NLP）、智能图像分析等领域的发展，邮储银行正研究开发一种基于多人对话的“U星”学习型智能答疑机器人。通过运用AI应用能力框架，引入“@”、“引用”等多人对话机制，实现了线上答疑流程创新，减少重复答疑并持续提高答疑准确性，结合智能交互、实时交互、图文交互，不断强化答疑能力，助力业务发展。

[关键词]多人对话;答疑;自然语言处理

[中图分类号]TP242.6 [文献标志码]A [文章编号]2095–6487（2021）11–0–02

Research and Develop a Learning Intelligent Answering

Robot Based on Multi-Person Dialogue

Zhu Zhe-ren， Liu Min， Chen Peng

[Abstract]The banking industry is a typical knowledge-intensive field. Due to factors such as the continuous updating of business rules and systems， and the flow of personnel at all levels， various departments have generated the need for knowledge reuse for business answering. With the development of deep learning and machine learning in natural language processing （NLP）， intelligent image analysis and other fields， Postal Savings Bank is researching and developing a "U-star" learning intelligent answering robot based on multi-person dialogue. Through the use of the AI application capability framework and the introduction of multi-person dialogue mechanisms such as “@” and “quotation”， the innovation of online Q&A process is realized， repeated Q&A is reduced and the accuracy of Q&A is continuously improved， combined with intelligent interaction， real-time interaction， and image-text interaction， Continuously strengthen the ability to answer questions and help business development.

[Keywords]multi-person dialogue; question answering; natural language processing

1 將机器人引入多人对话答疑的背景

银行业是典型的知识密集领域，大型商业银行往往具有庞大的组织结构、复杂的系统功能以及丰富的业务产品种类。同时，各级机构人员时常进行岗位轮换，需要学习各项业务、产品或系统应用。随着行内各项业务规则及系统功能不断更新，银行业从业人员内部的知识共享、复用等需要知识服务的场景极为多见。根据调研反馈，为增强业务学习能力，邮储银行总行业务专家需要在即时通信软件建立各种沟通群，花费大量时间精力为各级分支机构提供答疑，沟通效率和效果亟待提升。

近年来，以深度学习为代表的新一代人工智能技术极大地推进了产业界技术革新，同时重塑了银行业的业务模式。深度学习、机器学习方法在自然语言处理、智能图像分析等领域取得了长足进步，进而也带动了信息技术的发展。……