期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2019
卷号：97
期号：4
页码：1267-1277
出版社：Journal of Theoretical and Applied
摘要：Gunadarma University has been developing a Speech Intelligence Bot used either simply to automate exchanging information or as a better customer services in many industrial needs. For a basic use, this bot should be able to accept people�s oral questions and answer them appropriately. This conversation bot is using Sphinx-4 as its base technology and the process consist of human sending speech to speech recognition system and convert it into text, then the text is analyzed with MegaHal Algorithm supported with a knowledge base data, then generates reponse in a text format and convert it into speech that is sent to human. Thus, this research purpose is to develop and analyze the speech conversation bot accuracy with an Automated Logging System, though the analyzation is still attempted manually. Before answering any question, the conversation bot must define sets of arranged words as grammar using JSGF Language to understand any word it might listen. This research has done a test of asking 20 questions with 3 times repetition on each question to accumulate accuracy scores on the recognition system itself and the bot conversation as application. And the result of the test shows that from all generically 90 questions in total was tested on the bot, it shown 90% confidence score in its recognition level. And for the conversation, the bot showed a considerably low confidence score with 60% from all recognized questions are answered correctly. This results might be affected by the lack of environment control, and logic perfection in how the system accepts voices and converts them into processable data, it needs many improvements in speech recognition. Also, a better and faster way to analyze the log data so any further research could achieve more accurate result is suggested.
