Discriminative Analysis of Linguistic Features for Typological Study

Discriminative Analysis of Linguistic Features for Typological Study Hiroya Takamura, Ryo Nagata, Yoshifumi Kawasaki Tokyo Institute of Technology, Konan University, Sophia University takamura@pi.titech.ac.jp, nagata-acl@hyogo-u.ac.jp, kyossii@gmail.com Abstract We address the task of automatically estimating the missing values of linguistic features by making use of the fact that some linguistic features in typological databases are informative to each other. The questions to address in this work are (i) how much predictive power do features have on the value of another feature? (ii) to what extent can we attribute this predictive power to genealogical or areal factors, as opposed to being provided by tendencies or implicational universals? To address these questions, we conduct a discriminative or predictive analysis on the typological database. Specifically, we use a machine-learning classifier to estimate the value of each feature of each language using the values of the other features, under different choices of training data: all the other languages, or all the other languages except for the ones having the same origin or area with the target language. Keywords: language typology, linguistic feature, WALS, machine learning 1. Introduction There are numerous languages in the world. They are char- acterized from various viewpoints including the vocabu- lary, the syntactic rules, and the pronunciation system. In the language typology, the characteristics of languages are used to discuss the classification of languages and the sim- ilarity or dissimilarity between languages. The characteris- tics (or features1 henceforth) are the backbone of the lan- guage typology. Part of the findings with regard to the linguistic features is aggregated as databases. One of the largest databases is the World Atlas of Language Structures (WALS) (Dryer and Haspelmath, 2014). WALS encom- passes a wide range of linguistic features together with their values for various languages. We would like to note that some linguistic features are sometimes informative to each other. There are believed to be tendencies or implicational universals between fea- tures (Comrie, 1981). For example, it is widely known that, if VSO is the dominant order of a language, then the lan- guage has prepositions (Universal 3 by Greenberg (1963)). In other words, the values of features provide clues to the value of another feature. This fact brings up the following two questions: (i) how much predictive power do features have on the value of another feature? (ii) to what extent can we attribute this predictive power to genealogical or areal factors, as opposed to being provided by tendencies or implicational universals? To address these questions, we conduct a discriminative or predictive analysis on the typological database. Specifi- cally, we use a machine-learning classifier to estimate the value of each feature of each language using the values of the other features in WALS, under different choices of training data: all the other languages, or all the other lan- guages except for the ones having the same origin or area with the target language. 1To avoid confusion with feature of machine learning as in feature vector, in this paper, we use the term attribute …

Content truncated to 500 words