2015 Indonesian Syllabification
2014 Indonesian Grapheme-to-Phoneme Conversion
2013 Language Technology
Kajian Information Retreival Guna Membangun Sistem Deteksi Indikasi Plagiat, Agung Toto Wibowo, Suyanto, Ari M. Barmawi. Penelitian Hibah Bersaing yang dibiayai sebesar Rp 96.080.000 oleh Direktorat Jenderal Pendidikan Tinggi, Kementerian Pendidikan Nasional selama dua tahun dengan Surat Perjanjian Pelaksanaan Penugasan Penelitian Hibah Bersaing Nomor: 0926/K4/KL/2013
Segmentation of Speech into Syllable Units for Indonesian Language. Frame the given speech, calculate the energy rate of each frame, smooth the energy using fuzzy logic, and then define the boundaries of syllable segments.
Neural-Based Word-to-Syllable Decomposition for Spoken Indonesia Language. First, convert grapheme to phoneme so that a symbol states one phoneme. Then, split the phonemic symbol of word into syllable based on the syllable patterns trained to the neural network.
New design of Indonesian LVCSR Using Combined Phoneme and Syllable Models. Most recent large vocabulary continuous speech recognition (LVCSR) systems are based on context-dependent phonemes (usually tri- or quinta-phone) with a frame-based approach. This approach usually suffers from a lack of robustness againts background noise. This paper discusses a new design of Indonesian LVCSR based on critical study on some recent papers related to both phoneme-based and syllable-based speech recognition systems. The study concludes that Indonesian LVCSR is better to be designed using combined phoneme and syllable models, where a segment containing either consecutive consonants or consecutive vowels is processed using syllable model. Other segment is treated as phoneme using a frame-based approach. However, some problems related to the quite large number of syllables should be properly managed.
Automatic Speech Recognition-Based Information Center for Indonesian Language. This reserach focuses on implementation of CMU Sphinx-4 in an automatic speech recognition-based information center (ASRIC) for Indonesian language. The ASRIC designed to handle more than 20 thousands directory services as in the yellow pages in Bandung. It uses a vector space model (VSM) to improve the performance of statistical language model and to recognize the user utterance more flexibly. Testing to an Indonesian speaker shows that VSM is capable of reducing the query error rate, but the response time is quite low. The VSM also make ASRIC more flexible to recognize spoken sentences of users which misses some keywords, contains random ordered words, or consists of unimportant words.
2012 Language Technology
Kajian Information Retreival Guna Membangun Sistem Deteksi Indikasi Plagiat, Agung Toto Wibowo, Suyanto, Ari M. Barmawi. Penelitian Hibah Bersaing yang dibiayai sebesar Rp 96.080.000 oleh Direktorat Jenderal Pendidikan Tinggi, Kementerian Pendidikan Nasional selama dua tahun dengan Surat Perjanjian Pelaksanaan Penugasan Penelitian Hibah Bersaing Nomor: 0926/K4/KL/2013
IndoVMS: Indonesian Voice Messaging Service. This is an applied reserach to develop a mobile application called Yooi (an Indonesian slang word that means OK in English). Yooi enable you to dictate and send an SMS using your voice. It provides formal Indonesian language, slang, and three local languages: Javanese, Sundanese, Batak. Yooi is multi speaker so that can be used for teenagers or elderly people and different dialects. A paper for Yooi was presented in The 5th International Conference on Interaction Sciences: IT, Human and Digital Content (ICIS 2012), Jeju island, South Korea, on 26-28 June 2012.
Indonesian LVCSR for general domain. This research addresses new design of LVCSR for Indonesian language. Explore so many theory and think of the best strategy to map many challenges in Indonesian language.
IndoEMR: Indonesian Electronic Medical Record. Drug allergy detection and Drug interaction detection.
2011 Language Technology
Wikiphone: voice search for Indonesian language
Indonesian ASR for specific domain
Indonesian LVCSR for general domain
IndoPDS: Plagiarism Detection System in Indonesian Documents. Based on learning techniques. Designed for Indonesian documents
Supply Chain Management. Optimization based on evolutionary computation approach. Designed to run on cloud computing
University Timetabling Systems for Course, Student, and Examination. Based on evolutionary computation approach. Designed to be general for any university. Simple for any user. Real time.
IndoEMR: Indonesian Electronic Medical Record. Drug allergy detection. Drug interaction detection.
Theory of Evolutionary Computation. Globally Evolved Dynamic Bee Colony Optimization. Evolutionary Discrete Firefly Algorithm for Travelling Salesman Problem.
2010 Indonesian LVCSR
Validating Indonesian Speech Corpus of 400 speakers
Manual labeling for some speakers
Develop a Language Model using 500 thousand sentences
Develop a dictation system
2008-2009 University Course Timetabling (Sistem Penjadwalan Perkuliahan)
Lecture timetabling without any conflict for lecturer and class (Membangun Jadwal Perkuliahan sampai Level Kelas Tidak ada bentrok jadwal dosen maupun kelas)
Student timetabling to minimize number of conflict for students (Membangun Jadwal Perkuliahan sampai Level Mahasiswa (Meminimasi bentrok jadwal mahasiswa)
2007 – 2008
Indonesian Grapheme to Phoneme (IndoG2P). Convert Grapheme to Phoneme. Decision Tree Learning.
2005 – 2006
Signal Energy-Based Automatic Speech Splitter in Developing Speech Corpus
Modified Least-to-Most Greedy Algorithm to Search a Minimum Sentence Set. Modified LTM Greedy Search.
Development of Indonesian Text and Speech Corpus for Large Vocabulary Continuous Speech Recognition. Phonetical balance sentence set. 400 speakers. 4 dialects: Javanese, Sundanese, Batak, Betawi.