Rowena Cristina L. Guevara, Arvin Jeremy C. Agoncillo,
Mio Miguel O. Galang, Kristina R. Tejerero,
Prospero C. Naval Jr., Ed Peter G. Cabalfin,
and Ricardo Ma. D. Nolasco
(Presented at the 1st Philippine Conference-Workshop on Mother Tongue-based Multilingual Education held at the Capitol University, Cagayan de Oro City, Philippines, on Feb. 18-20, 2010.)
ABSTRACT: In this paper, we describe the different databases for Filipino speech and Filipino sign language developed by the University of the Philippines – Diliman. The Filipino Speech Corpus (FSC) was created by the Digital Signal Processing Laboratory (DSP Lab) of the UP Electrical and Electronics Engineering Institute for the purpose of Filipino speech recognition. The FSC contains text and spontaneous speech of 100 Filipino speakers from all over the country. A second database, FSC 2, was developed for diphone-based Filipino speech synthesis. FSC 2 consists of 4 speakers. A Filipino Sign Language (FSL) database was developed by the Computer Vision and Machine Intelligence Group of the UP Department of Computer Science and the DSP Lab. The FSL is a video recording of 117 unique FSL signs performed by three native FSL signers. Only traditional signs were used. The FSL was developed for sign language recognition.
Through the databases mentioned, the standardization of Filipino on the basis of actual usage could be achieved despite its rapid and dynamic development. For example, FSC has been used to standardize the spelling of English loan words in the Filipino dictionary through vowel mapping from the English language to their Filipino equivalents. Finally, the FSL aims to unify the deaf communities in the country through the use of a national set of signs.
The University of the Philippines in Diliman is currently developing databases that are geared towards the development of Filipino, our national language. Speech and sign language databases were created to help standardize Filipino both as a spoken language and as a signed language.
There had been a lot of steps made in order to establish our “national language”. In 1936, the newly-established Surian ng Wikang Pambansa (National Language Institute) chose Tagalog as the basis of the “national language” primarily because it is widely spoken and widely understood in most parts of the country. In 1959, this “national language” was named “Pilipino”, and later renamed “Filipino” through the 1973 Constitution. The 1987 Constitution established provisions in using Filipino as the medium of official communication and medium of instruction in educational institutions .
Despite the government’s efforts to create such a policy, the Philippine society continues to be multilingual. Filipino may be the national language, but English is used in most official and written communication . In education, literacy development in children’s formative years is done usually through the vernacular or an ethnic home language. Transitions occur in elementary and high school, where the media of instruction becomes English and Filipino. Finally, English is used in almost all subjects and all levels of tertiary instruction, as well as in other forms of scholarly discourse , although oral intellectualized discourse in campus is done by mixing and switching between English and Filipino.
Deeper understanding and mastery of a language lies in the proper and constant use of the said language. With English dominating most discourses, especially in the academe, the development of local languages and Filipino as languages of intellectual discourse becomes a challenge. The Filipino Speech Corpus (FSC) can be used to help in the instruction of Filipino as spoken and written by a variety of users, and ultimately in the development of our national language.
Similarly, the diversity of sign languages used in the Philippines calls for development of a standard set of signs. The American Sign Language and indigenous signs contribute to the wide array of traditional and emerging signs used in our country . The Filipino Sign Language Database is targeted towards collating all the signs used in the Philippines for development of instructional materials and technologies for the deaf community.
2.1 Filipino Speech Corpus
The Filipino Speech Corpus (FSC) was created by the Digital Signal Processing Laboratory (DSP Lab) of the UP Electrical and Electronics Engineering Institute with the help of the UP Department of Linguistics in 2002. The corpus contains more than 100 hours of read Filipino text and spontaneous Filipino speech recorded from 50 female and 50 male speakers from different parts of the country using high-fidelity audio recording equipment. The recorded speech was then segmented manually into sentences, words and phonemes . A variation of the FSC was created in 2009 that included diphones – a pair of phones that illustrates the transition between the two phones . The recent version is now being used in various Filipino speech synthesis and analysis research.
2.2 Filipino Sign Language
The Computer Vision and Machine Intelligence Group of the UP Department of Computer Science and the DSP Lab developed the Filipino Sign Language (FSL) Database. It is a compilation of video recordings of traditional signs used in the Philippines. Three native FSL signers were recorded individually, each performing 117 unique FSL signs. The collected video data were then pre-processed and edited to facilitate development of a Filipino sign language recognition technology .
3. DATABASES: THEIR USES
3.1 Current and Ongoing Projects
Since its creation, the FSC and the FSL have been used in several speech and language-based research projects.
3.1.1 Real-time Implementation of a Low Bit Rate Filipino Speech Codec Using Hidden Markov Model-based Speech Recognition/ Synthesis
A speech recognition-synthesizer system specific for Filipino has been developed using the FSC. Different mathematical techniques were used and are currently being used to further improve the features of the system. Eventually, the software will be extended to hardware implementation .
3.1.2 Symbian OS S60 Platform Implementation of a Smartphone-based Automatic Speech Recognition System for the Filipino Speech Corpus
A mobile phone-based speech-to-text application is being developed further with the help of the FSC. The mobile phone receives voice input, which is then sent to a personal computer via Bluetooth for analysis and conversion to text. Finally, the computer sends the generated text back to the mobile phone for display . This technology will eventually contribute to a more convenient array of voice-activated mobile communication devices.
3.1.3 Prosody Development for Filipino TTS System
Prosody refers to the characteristics of speech such as rhythm, stress and intonation. A text-to-speech system aimed at producing natural-sounding synthesized speech is being developed using the FSC. In synthesis, phones lifted from the corpus are put together. Prosody-varying algorithms are then used to make the synthesized speech sound more like a native Filipino speaker .
3.1.4 Towards the Standardization of the Filipino Language: Focus on the Vowels of English Loan Words
Loan words are words borrowed from another language with little or no change in the spelling and pronunciation. Vowels of the loan words from the English language were mapped onto the Filipino vowels in the FSC. On the basis of such mapping, the spelling of English loan words is standardized in the Filipino dictionary .
3.1.5 Filipino Sign Language Recognition
The FSL was of great help in creating a system that detects and classifies Filipino signs. Recorded videos of signers have been processed using different mathematical methods to develop recognition algorithms .
3.2 Proposed Applications of Databases to Educational Institutions
Present technological advances have made teaching methods that were unthinkable years ago become possible. The FSC and the FSL are envisioned to be the backbone of future electronic educational tools that would assist teachers in promoting a better learning experience for their students. These tools could never take the place of a human teacher. Instead, these tools could help contribute to a whole new level of learning. A possible use of the FSC and the FSL is in distance education. Synthesized words from a specific language may accompany the books that a home-schooled student usually has. Sets of vocabulary words in the form of audio recordings may be given to the students as listening assignments. These listening assignments may go together with the reading lists that a student usually has to finish. Moreover, audio-books of Filipino literature generated from synthesized speech may also be given to the students to supplement their understanding while enjoying our colorful local literature. With the help of the FSC and the FSL, virtual learning environment websites, language exchange websites and language portals may be developed. A virtual learning environment (VLE) is a website where electronic resources may be uploaded and downloaded. VLEs dedicated to Filipino language learning can contain synthesized Filipino vocabulary words, as well as video demonstrations of Filipino signs, all of which can be downloaded for self-study. Language exchange websites and language portals, on the other hand, are websites where people who want to learn a specific language sign up and initiate interaction with people who share the same passion for learning. Incorporated with materials developed from the FSC and the FSL, learning Filipino with Internet users from different parts of the world is possible.
Software for computer-assisted learning may be also developed with the FSC and the FSL. Spoken Filipino synthesized with the help of the FSC can be made as a basis for evaluating a student’s oral communication skills. When a student learns new words, pronunciation meters would help them compare how they spoke a specific word with the correct pronunciation. Grammar tutors may supplement the discussion of teachers regarding a specific language’s grammar. Evaluative measures like practical exams can also take the form of a mock conversation with a computer. His grammar and conversational skills can be gauged as he carries a situational conversation with a computer that can speak with the help of the FSC.
Development of multimedia which could help in language learning both inside and outside the classroom can also be made possible with the databases. The research on Filipino sign language recognition with the help of the FSL may actually pave the way for the development of sign-to-speech/ text and speech/ text-to-sign software, which could really help the visually- and speaking-impaired. A machine records the deaf person’s gestures and converts it to the corresponding speech or text equivalent. Finally, younger deaf students could enjoy learning new signs with the help of animated characters created through a system wherein gestures are recognized and converted to animation.
The FSC may be expanded to accommodate other local languages that have no existing corpora. With that, multilingual audio dictionaries can be created, wherein a specific English word is presented in audio format with its corresponding equivalents in Tagalog, Cebuano, Hiligaynon, Bikolano, Ilokano and other indigenous languages and dialects. Ultimately, the extension of databases to other languages of the Philippine supports mother tongue-based multilingual education. These multilingual databases widen and deepen our understanding of how Philippine languages work, and provide us with lasting, multipurpose repositories of our languages that can be made available to interested researchers and users. They can also make analyses accountable and claims verifiable especially in cases where primary data in a particular language remain inaccessible.
The Filipino Speech Corpus (FSC) and the Filipino Sign Language (FSL) databases created by the University of the Philippines were initially geared towards developing technologies for speech synthesis, recognition and sign language recognition and processing. This paper has presented ways in which these databases can be used also to augment teaching. Specific applications were enumerated that would induce an environment more conducive and effective for learning. These applications are envisioned to help both the teachers and the students. Moreover, database development techniques can be used in creating databases for other languages in the country. Consequently, the applications suggested can be extended beyond Filipino and English, to all the languages used in the Philippines.
We would like to thank the Office of the Vice-President for Academic Affairs and the Office of the Vice-chancellor for Research and Development, for without them, these projects would not be realized.
 Ang, F. M. (2007). Symbian OS S60 Platform Implementation of a Smartphone-based Automatic Speech Recognition System for the Filipino Speech Corpus. University of the Philippines – Diliman: Electrical and Electronics Engineering Institute.
 Cabalfin, E. et al. (2008). Filipino Sign Language Recognition. University of the Philippines – Diliman: Department of Computer Science.
 Gonzales, A. Language Planning in Multilingual Countries: The Case of the Philippines. De La Salle University – Manila.
 Guevara, R. et al. (2002). Development of a Filipino Speech Corpus. University of the Philippines – Diliman: Electrical and Electronics Engineering Institute.
 Lazaro, L. R. (2009). Incorporating Duration and Intonation Models in Filipino Speech Synthesis. University of the Philippines – Diliman: Electrical and Electronics Engineering Institute
 Rara, K. et al. (2008). Towards the Standardization of the Filipino Language: Focus on the Vowels of English Loan Words. University of the Philippines – Diliman.
 Unknown. (2010). The Official Government Portal of the Republic of the Philippines. http://www.gov.ph/