Weather Forecast Voice System

Weather Forecast Voice System Thanh Do Faculty of Information Technology, Ho Chi Minh City University of Foreign Languages and Information Technology (HUFLIT) dodinhthanh@huflit.edu.vn Abstract This paper introduces Weather Voice System applied to information commands-responses system through Public Switched Telephone Network (PSTN). Being able to understand users' voice commands, our system would assist them during their searching weather forecast in Vietnamese towns and cities by communicating in Vietnamese through Telephone Network. What makes this system different from others consists of its ability to analyze syntax and semantics of voice commands after their being recognized by speech recognizer component. Weather Voice is composed of main components that are there: 1) Telephone network communicator, 2) Vietnamese speech recognizer, 3) Natural language processor, and 4) Vietnamese speech synthesizer. To the best of our knowledge, this is one of the first systems in Vietnam to combine text language processing and speech processing. This event could make voice applications more intelligent with aim to communicate with humans in natural language at high precision and at quick speed. The fact that experiments have proved our expectation as well as the friendliness of the system to users reveals the practicality of our research. Keywords: Voice Application, Natural Language Processing, Voice Server, WeatherVoice 1. Introduction In the recent years, Vietnamese voice processing researches have found themselves making significant progress. Among them, two which must be mentioned are Voice recognizing (transforming audible signals into visible text) and Voice synthesizing (transforming visible text into audible signals), their authors, Information Technology Institute (Vietnamese Science and Technology Institute) and Natural Science University (Vietnamese National University), have The 2018 Conference on Computational Linguistics and Speech Processing ROCLING 2018, pp. 126-139 ©The Association for Computational Linguistics and Chinese Language Processing 126 conducted and published many works high estimated [2],[4],[6],[8],[14]. Especially in 2012, the authors from Natural Science University succeeded in building Voice Server System [9] to explore potential applications of Voice Processing whose use was made of by media systems, such as Voice Information Providing System through mobile and fixed phones, this work was awarded the second prize of Vietnamese Gifted. However, the works above only focused on improving the processing of Vietnamese voice and still neglected the semantic analyzing of voice commands. Weather Voice is Voice Server System built on the combination of speech language processing and text language one. It can recognize numerous patterns of Vietnamese voice commands in order to transform them into text, then analyze their syntax and semantics, afterwards carry out searching data base, and eventually answer to users with the data searched in Vietnamese voice. The analyzing syntax and semantics of commands is assisted by DCG (Definite Clause Grammar) [3],[7],[11],[12],[17]. During the voice processing, we call for HTK (Hidden Markov Model Toolkit) [13] to recognize voices and apply Unit - selection [1] method to synthesizing voices. 2. System structure The system is designed to achieve the following functions: Recognizing query statements through phones; analyzing them; searching data from database; answering to users through phones. Here is the scenario in …

Content truncated to 500 words