Browse Prior Art Database

System and Method for Reinforced Resource Selection for Conversational Systems

IP.com Disclosure Number: IPCOM000249153D
Publication Date: 2017-Feb-08
Document File: 3 page(s) / 78K

Publishing Venue

The IP.com Prior Art Database

Abstract

The usage of big data has brought significant improvement in statistical text understanding, however, as the different textual resources differ much in quality (e.g., professional documents, amateur documents, and forum discussion), selecting high-quality textual resources is critical for conversational systems. This article proposes to use reinforcement learning approach to select high-quality textual resources based on the runtime usage of these resources.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 72% of the total text.

1

System and Method for Reinforced Resource Selection for Conversational Systems

The usage of big data has brought significant improvement in statistical text understanding, however, as the different textual resources differ much in quality (e.g., professional documents, amateur documents, and forum discussion), selecting high-quality textual resources is critical for conversational systems.

Figure 1 shows the how the low textual resource quality may affect the runtime conversations. In many cases, the data quality affects the resultant machine learning algorithms for many modules in building cognitive conversational systems. The modules that use machine learning algorithms include Natual Language Classifier (NLC), Named Entity Recognition (NER), Emotion Detection, etc. Imprecise, erroneous, misleading texts may cause faulty classifiers, entities, and emotions, which will cause faulty understanding and handling in conversational system. Finally, it will lead to faulty conversations percevied by the user.

2

Figure 1. Faulty conversations due to the low textual resource quality.

To overcome such issues, this articles proposes a reinforced knowledge mining framework. This framework applies Reinforcement Learning (RL) to help exclude imprecise or erroneous resources through the interactions with end users.

Figure 2 shows the components of the system. The reinforcement learning for resource selection will guide the resource selection process each time. The resource selection is s...