Browse Prior Art Database

Method and Apparatus to support auto-switch from virtual agent to human agent in call center

IP.com Disclosure Number: IPCOM000249109D
Publication Date: 2017-Feb-07
Document File: 4 page(s) / 82K

Publishing Venue

The IP.com Prior Art Database

Abstract

In virtual agent system, how to automatically seamlessly switch an ongoing conversation with client from a virtual agent (VA) to a human agent (HA) is the key problem to be solved by this disclosure. the disclosure presents a solution where to consider the problem of virtual agent to human agent (VA-to-HA) as a prediction problem-based on the historical conversation, to predict if a VA-to-HA event should happen. If yes, do the switch, otherwise, continue to dialog with VA. The prediction model is realezed as a binary classification problem in which the features can be emotion, emotion change, conversation atmosphere, conversation relevance, previous conversation experience and so on.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

1

Method and Apparatus to support auto-switch from virtual agent to human agent in call center

Background: Virtual agent is emerging as a promising technology to improve the business performance and also reduce the labor cost at client centers/call centers. The core technology of virtual agent is the bot/conversational system. However, the reality of current bot technology is we cannot well handle all conversations. Human intervention is a MUST, but missed now.

Problem: How to automatically seamlessly switch an ongoing conversation with client from a virtual agent (VA) to a human agent (HA) is the key problem to be solved by this disclosure. The key is when to make the switching happen.

Main idea of the disclosure: Consider the problem of virtual agent to human agent (VA-to-HA, or VA2HA) as a prediction problem - Based on the historical conversation, to predict if a VA-to-HA event should happen. If yes, do the switch, otherwise, continue to dialog with VA.

The following figure shows how a bot system works with VA-to-HA technology. The End user interacts with bot system via conversation channels, such as, phone, wechat etc. the user's utterance is the input for a bot, the bot's utterance is the output of the bot system. Conversation orchestrator received the input content and firstly do the Natural Language Processing (NLP) / Natural Language Understanding (NLU) analysis, meanwhile the content is also processed by persona management, context management, emotion analysis and VA-to-HA (VA2HA) evaluator. Persona management is to...