Spoken-Based Sentence Retrieval System (Case Study: Standard Operating Procedure on Informatics ITB)

• Main Author: LAKSMANA PRAMUDITA (NIM : 13511042), BIMA
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/21435
• Institution: Institut Teknologi Bandung

Standard Operating Procedure or SOP from an institution must be a well-known thing by all stakeholders in the institution. In reality, the flow of information that going through our daily life in university environment is certainly huge. SOPs were designed to be systematic making it easier for everyone to know the steps of each procedure. In the case of accessing information it is better to create a system that can handle the delivery of SOP information in order to speed up the delivery of the information thoroughly. <br /> <br /> <br /> <br /> Spoken-based Sentence Retrieval System is built using an integrated system of 2 large modules, Automatic Speech Recognition or ASR and Sentence Retrieval. The system process begins with voice input to be sent to the server to be processed into a text. The development of the ASR module is done using the Hidden Markov Model and N-gram method. Dam the development of the Sentence Retrieval module is done by utilizing Information Retrieval method or IR using Vector Space Model TFIDF as ranking modeling for SOP text file. <br /> <br /> <br /> <br /> Results of the system testing in this final project has fulfilled the functionality by recognizing the user’s voice and change them into text form in the ASR module with an accuracy level of 70%. The system has also been able to recognize and handle relatable questions from users with 100% success rate. Nevertheless, weaknesses in the system can still be found inside user’s voice recognizer in which still could be studied further in this final project.