Plan Optimization to Bilingual Dictionary Induction for Low-resource Language Families

Creating bilingual dictionary is the first crucial step in enriching low-resource languages. Especially for closely related ones, it has been shown that constraint-based approach useful inducing lexicons from two dictionaries via pivot language. However, if there are no available machine-readable as input, we need to consider manual creation by native speakers. To reach a goal of comprehensively create multiple dictionaries, even already have several existing still difficult determine execution order reducing total cost. Plan optimization composing with consideration methods and their costs. We formalize plan creating utilizing Markov Decision Process (MDP) get more accurate estimation most feasible optimal least cost before fully implementing lexicon induction. model prior beta distribution induction precision language similarity polysemy topology $\beta$?> parameters. It further used function state transition probability. estimated all investment plans baseline evaluating proposed MDP-based an evaluation metric. After posterior batch experiments construct second experiments, result shows 61.5% reduction compared 39.4% MDP plan. The proposal outperformed on