The Annotation of Antonym Information in the ‘Word List by Semantic Principles’

The Annotation of Antonym Information in the ‘Word List by Semantic Principles’ Sachi Kato Mejiro University Masayuki Asahara NINJAL, Japan masayu-a@ninjal.ac.jp Nanami Moriyama NINJAL, Japan Asami Ogiwara NINJAL, Japan Makoto Yamazaki NINJAL, Japan Abstract This article presents an antonym database con- struction based on the Word List by Seman- tic Principles. Antonym word pair candidates were extracted, after which we performed a cognitive experiment to rate the antonyms ac- cording to crowdsourcing. We then annotated the types of antonyms. The statistics based on the rates and types of antonyms indicated that closed antonyms tend to be considered ‘antonyms’ rather than open antonyms. 1 Introduction This article presents an antonym database construc- tion based on the ‘Word List by Semantic Principles’ (WLSP) (Kokuritsu Kokugo Kenkyusho, 2004). The National Institute for Japanese Language and Linguistics (NINJAL) has been developing language resources based on the WLSP, which comprises a collection of words that are classified and arranged by their meanings. The first version of the WLSP was published in 1964, and the revised and enhanced version was published in 2004. The database version of the WLSP was constructed by incorporating the contents of the revised and enlarged edition of the WLSP book. It was created in the CSV file format, and the total number of records was 101,070. The data can be accessed via the Internet1. Table 1 outlines the structure of the WLSP, and the words in the WLSP are assigned five digits that compose the article number. Further, the en- tries in the ‘Paragraph number etc.’ column in- 1https://github.com/masayu-a/WLSP clude three hyphenated numbers: paragraph num- bers, small paragraph number, and word number. These numbers indicate the hierarchical clusters of words. The article numbers indicate the syntactic cate- gory of ‘class’ and the several hierarchical seman- tic levels of ‘division’, ‘section’ and ‘article’. The categories are indicated with a one-integer digit to the left of a radix point and four fractional dig- its to the right of the radix point. An example is the word ホステス (‘hostess’), which is assigned a value of 1.2220. Here, the first digit ‘1’ desig- nates the syntactic part, which is termed the ‘Nomi- nal Word’, while ‘2220’ designates the hierarchical semantic parts. The first digit ‘.2’ denotes the top- level semantic category of ‘Subject’; the two dig- its ‘.22’ denote the second-level semantic category of ‘Companion’; and the four digits ‘.2220’ denote the finest-grained semantic category of ‘Host and Guest’. These five digits are thus referred to as the ‘WLSP number’. The syntactic categories include the following: 1. Nominal Word, 2. Verbal Word, 3. Modifier Word and 4. Other (e.g., Conjunction, Interjection, Greeting). The WLSP semantic-label hierarchy assigns the same label to antonyms and opposites. For example, paragraph 1.2220-2 in Table 1 includes the antonym pairs ホステス ‘hostess’ ⇔ ホスト ‘host’ by the type of 相補 (complementation). However, ゲスト ‘guest’ is also included in paragraph 1.5300-3, and it can be considered an antonym of ホステス and ホ スト by the type of 視点 (viewpoint). Therefore, it …

Content truncated to 500 words