A Development of Speech-based Restaurant Recommendation System for Nakhon Sawan Province

Abstract

Nakhon Sawan province provides various types of restaurants with diverse type of food and atmosphere. Costumers sometimes arrives at the restaurants, but some dishes have run out or found that the restaurants are closed. Some costumers do not know the route to the restaurants. In general, the users enter searching restaurant names using keyboard and sometimes they make typographical errors which consume their searching time. Therefore, this research aims to develop a speech-based restaurant recommendation system based on Android operating system for Nakhon Sawan province in order to reduce problems caused by using keyboards. The proposed speech-based restaurant recommendation system utilizes speech instead of a keyboard as an input and implements the system on Android mobile phones. The research methodology composed of five parts. First, the architecture of the proposed system. Second, storing restaurant information within the area of Nakhon Sawan province in MySQL database using PHP-based web application. Third, users can use their speech as searching input to the proposed system. JAVA then transformed a speech input to a text message. Word segmentation, deleting propositions and meaningless words are applied to the transformed text message and search process will be executed. Forth, the proposed system delivers the restaurant information to users as Thai speech messages using Google Translate TTS. Last, the route to the user’s request restaurant was displayed on Google map. The speech-based restaurant recommendation system also provides the restaurants information nearby the user’s geographical location and office hours in real time. The experimental results show that the average accuracy of the proposed system is 91.60% in the noiseless environment. The average accuracy in the noisy environment is decreased to 84.80%.