Mispronunciation Detection and Diagnosis Through a Chatbot

Mispronunciation Detection and Diagnosis Through a Chatbot

Marcos E. Martinez, Francisco López-Orozco, Karla Olmos-Sánchez, Julia Patricia Sánchez-Solís
DOI: 10.4018/978-1-7998-4730-4.ch002
(Individual Chapters)
No Current Special Offers


The interaction between humans and machines has evolved; thus, the idea of being able to communicate with computers as we usually do with other people is becoming increasingly closer to coming true. Nowadays, it is common to come across intelligent systems named chatbots, which allow people to communicate by using natural language to hold conversations related to a specific domain. Chatbots have gained popularity in different kinds of sectors, such as customer service, marketing, sales, e-commerce, e-learning, travel, and even in education itself. This chapter aims to present a chatbot-based approach to learning English as a second language by using computer-assisted language learning systems.
Chapter Preview


During day-to-day activities, human beings make use of natural language. Something that characterizes natural language is its ambiguity, especially when it is expressed in written format. Hence, Artificial Intelligence (AI) community has been extensively researched and developed techniques, algorithms, and tools in order to improve the human-computer interaction. Natural Language Processing (NLP) arises in 1950 from the need to interact with computers in a natural manner and extract information from human speeches (Kamath, Liu, & Whitaker 2019). Since then, various computer applications have been developed to accomplish this task. An example of these applications is chatbots, which are computer programs that use NLP to imitate human behavior to the point of making them believe that they are interacting with other humans (Roos, 2018). Chatbots are frequently used in areas such as customer service, marketing, sales, e-Commerce, travel, and even in education (Zumstein & Hundertmark, 2018).

The advantage of implementing a chatbot is that it allows users to access services at any time. In addition, the implementation of chatbots in education is of great potential since they can be appropriate partners for learning (Shawar, 2017). Also, they can be used to learn a second language through Computer Assisted Language Learning (CALL) systems. Pronunciation is usually the skill that causes most problems during second language learning. Diverse factors such as shyness, lack of practice, or bad feedback can affect the development of this skill. Computer-Assisted Pronunciation Training (CAPT) is a CALL sub-area which aims to help develop speech skills. These systems use Automatic Speech Recognition (ASR) and Mispronunciation Detection and Diagnosis (MDD) algorithms to detect pronunciation errors. Equip a chatbot with an MDD algorithm could emulate a second language teacher that provides pronunciation error feedback. Currently, there are some systems with these characteristics; however, they are focused on teaching vocabulary in isolated words (Heil, Wu, Lee, & Schmidt, 2016), depriving students’ opportunities to create their own sentences.

The aim of this chapter is to introduce a proposal integrating automatic speech recognition, chatbots, and mispronunciation detection and diagnosis in order to improve second language learners’ pronunciation skills. The purpose of using a chatbot is to achieve a conversation with a character emulating a real person. The structure of this chapter is as follows. First, the Background section provides a review of the literature on NLP, as well as some of its applications and techniques. Second, the Main Focus section describes the interest of this research, which is to integrate a mispronunciation detection module in a Chatbot. Additionally, a description of Computer-Assisted Language Learning Systems and Computer-Assisted Pronunciation Training is presented. The Solutions section shows one of the possible algorithms that could be used in the development of this research proposal. Finally, future trends are described, and a discussion of the partial results obtained is presented.

Complete Chapter List

Search this Book: