Distinguishing Japanese Non-standard Usages from Standard Ones

Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2323–2328 Copenhagen, Denmark, September 7–11, 2017. c⃝2017 Association for Computational Linguistics Distinguishing Japanese Non-standard Usages from Standard Ones Tatsuya Aoki† Ryohei Sasano‡ Hiroya Takamura† Manabu Okumura† †Department of Information and Communications Engineering, Tokyo Institute of Technology ‡Graduate School of Informatics, Nagoya University {aoki,takamura,oku}@lr.pi.titech.ac.jp sasano@i.nagoya-u.ac.jp Abstract We focus on non-standard usages of com- mon words on social media. In the context of social media, words sometimes have other usages that are totally different from their original. In this study, we attempt to distinguish non-standard usages on social media from standard ones in an unsuper- vised manner. Our basic idea is that non- standardness can be measured by the in- consistency between the expected mean- ing of the target word and the given con- text. For this purpose, we use context em- beddings derived from word embeddings. Our experimental results show that the model leveraging the context embedding outperforms other methods and provide us with findings, for example, on how to con- struct context embeddings and which cor- pus to use. 1 Introduction On social media such as Twitter, we often find posts that are difficult to interpret without prior knowledge on non-standard usage of words. For example, consider the following Japanese sen- tence1: 鯖の 負担が mackerel-POSS load-NOM増える increase-PRS “The load on a mackerel increases”, which does not make sense given the standard us- ages for the words in the sentence. But here, mack- erel is a non-standard usage that means computer server. The entire sentence should be interpreted as “The load on the computer server increases”. 1This is interlinear-gloss text representation. POSS, NOM, PRS respectively represent the possessive case, nom- inative case, and present tense. The third line is the standard translation of the Japanese sentence. The Japanese word “鯖 (saba)” (i.e., mackerel) is used to mean computer server by Japanese com- puter geeks because saba happens to have a pro- nunciation that is similar to s¯ab¯a (i.e., computer server). When a word is used in a meaning that is different from its dictionary meaning, we call such a usage non-standard.2 Non-standard usages can be found in many lan- guages (Sboev, 2016). For example, the word “catfish” means a ray-finned fish as in a standard dictionary, but on social media, it can mean a person who pretends to be someone else in or- der to create a fake identity. Such non-standard usages would be an obstacle to a variety of lan- guage processings including machine translation; Google Translate cannot correctly interpret exam- ples such as this. Humans, however, would be able to notice non-standard usages from the inconsis- tency between the expected word meaning and the context. The purpose of this work is to develop a method for distinguishing non-standard usages of Japanese words from standard ones. Since it is im- practical to construct a large labeled data set for each word, we focus on unsupervised approaches. The main idea in our method is that the differ- ence …

Content truncated to 500 words