Designing a Collaborative Process to Create Bilingual Dictionaries of Indonesian Ethnic Languages

Designing a Collaborative Process to Create Bilingual Dictionaries of Indonesian Ethnic Languages 1Arbi Haza Nasution, 2Yohei Murakami, 3Toru Ishida 1,3Department of Social Informatics, Kyoto University, 2Unit of Design, Kyoto University Kyoto, Japan 1arbi@ai.soc.i.kyoto-u.ac.jp, 2yohei@i.kyoto-u.ac.jp, 3ishida@i.kyoto-u.ac.jp Abstract The constraint-based approach has been proven useful for inducing bilingual dictionary for closely-related low-resource languages. When we want to create multiple bilingual dictionaries linking several languages, we need to consider manual creation by a native speaker if there are no available machine-readable dictionaries are available as input. To overcome the difficulty in planning the creation of bilingual dictionaries, the consideration of various methods and costs, plan optimization is essential. Utilizing both constraint-based approach and plan optimizer, we design a collaborative process for creating 10 bilingual dictionaries from every combination of 5 languages, i.e., Indonesian, Malay, Minangkabau, Javanese, and Sundanese. We further design an online collaborative dictionary generation to bridge spatial gap between native speakers. We define a heuristic plan that only utilizes manual investment by the native speaker to evaluate our optimal plan with total cost as an evaluation metric. The optimal plan outperformed the heuristic plan with a 63.3% cost reduction. Keywords: Bilingual Dictionary Creation, Low-resource Languages, Closely-related Languages 1. Introduction Nowadays, machine-readable bilingual dictionaries are be- ing utilized in actual services (Ishida, 2011) to support intercultural collaboration (Ishida, 2016; Nasution et al., 2017b), but low-resource languages lack such sources. Ob- viously bilingual lexicon extraction is highly problematic for low-resource languages due to the paucity or outright omission of parallel and comparable corpora. We intro- duced the promising approach of treating pivot-based bilin- gual dictionary induction for low-resource languages as an optimization problem (Nasution et al., 2016; Nasution et al., 2017c) where bilingual dictionaries are the only lan- guage resource required. Despite the high potential of our approach in enriching low-resource languages, it faces nu- merous issues when trying to create plans to implement multiple bilingual dictionaries for a set of low-resource lan- guages like Indonesian ethnic languages. When actually implementing our constraint-based bilingual dictionary in- duction approach, we need to consider the inclusion of more traditional methods like manually creating the bilin- gual dictionaries by native speaker. In spite of the high cost, this will be unavoidable if no machine-readable dictionar- ies are available. Given the various methods and costs that may need to be considered, we recently introduced a plan optimizer to find the feasible optimal plan of creating multi- ple bilingual dictionaries with the least total cost (Nasution et al., 2017a). In this project, to create bilingual dictionary DA−B between ethnic language LA and ethnic language LB, there is also a difficulty in finding a bilingual native speaker of two ethnic languages. To overcome this limita- tion, we can firstly create triple TA−ID−B using the com- mon language, Indonesian as pivot language LID where SID−A, a native bilingual speaker of Indonesian language LID - ethnic language LA and SID−B, a native bilingual speaker of Indonesian language LID - ethnic language LB collaborate by explaining the senses with Indonesian lan- guage. …

Content truncated to 500 words