Towards the Rapid Development of a Natural Language Understanding Module

Abstract

When developing a conversational agent, there is often an urgent need to have a prototype available in order to test the application with real users. A Wizard of Oz is a possibility, but sometimes the agent should be simply deployed in the environment where it will be used. Here, the agent should be able to capture as many interactions as possible and to understand how people react to failure. In this paper, we focus on the rapid development of a natural language understanding module by non experts. Our approach follows the learning paradigm and sees the process of understanding natural language as a classification problem. We test our module with a conversational agent that answers questions in the art domain. Moreover, we show how our approach can be used by a natural language interface to a cinema database.

A Simultaneous Recognition Framework for the Spoken Language Understanding Module of Intelligent Personal Assistant Software on Smart Phones

“…In addition, we used the paired t-test and Wilcoxon singed rank test to verify statistically significant between our framework and compared baseline framework. The pipeline framework (Moreira et al, 2011) is used a baseline system in our experiments because it is the most common method for multidomains SLU module.…”

“…The performance of the SLU module is averaged on 5-fold cross validation. In addition, we used the paired t-test and Wilcoxon singed rank test to The pipeline framework (Moreira et al, 2011) is used a baseline system in our experiments because it is the most common method for multidomains SLU module.…”

Improving Classification-Based Natural Language Understanding with Non-Expert Annotation

“…(Core and Allen, 1997; Bunt et al, 2010), user information needs, e.g. (Moreira et al, 2011), or stand in for entire semantic frames, e.g. (DeVault and Traum, 2013).…”

“…Although data-driven approaches are now commonly used to address the NLU problem as one of classification, e.g. (Heintze et al, 2010;Leuski and Traum, 2010;Moreira et al, 2011), where input utterances are mapped automatically into system-specific categories, the dependence of such approaches on training data annotated with semantic classes or dialogue acts creates a chicken and egg problem: user utterances are needed to create the annotated training data necessary for NLU by classification, but these cannot be collected without a working system that users can interact with.…”

Question Generation based on Lexico-Syntactic Patterns Learned from the Web

“…Following this idea, the FalaComigo project invites tourists to interact with virtual agents through questions/answers and multiple-choice tests. 1 Agent Edgar (Figure 1) is the face of the project, currently operating in the Palace of Monserrate (Moreira et al 2011).…”