Sina Mandarin Alphabetical Words:A Web-driven Code-mixing Lexical Resource

Proceedings of the 1st Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 10th International Joint Conference on Natural Language Processing, pages 833–842 December 4 - 7, 2020. c⃝2020 Association for Computational Linguistics 833 Sina Mandarin Alphabetical Words: A Web-driven Code-mixing Lexical Resource Rong Xiang1, Mingyu Wan12, Qi Su2, Chu-Ren Huang1, Qin Lu1 1The Hong Kong Polytechnic University, 11 Yuk Choi Road, Hong Kong (China) csrxiang,csluqin@comp.polyu.edu.hk, churen.huang@polyu.edu.hk 2Peking University, 5 Yiheyuan Road, Beijing (China) wanmy, sukia@pku.edu.cn Abstract Mandarin Alphabetical Word (MAW) is one indispensable component of Modern Chinese that demonstrates unique code-mixing idiosyn- crasies influenced by language exchanges. Yet, this interesting phenomenon has not been prop- erly addressed and is mostly excluded from the Chinese language system. This paper ad- dresses the core problem of MAW identifica- tion and proposes to construct a large collec- tion of MAWs from Sina Weibo (SMAW) us- ing an automatic web-based technique which includes rule-based identification, informatics- based extraction, as well as Baidu search en- gine validation. A collection of 16,207 quali- fied SMAWs are obtained using this technique along with an annotated corpus of more than 200,000 sentences for linguistic research and applicable inquiries. 1 Introduction Mandarin Alphabetic Words (MAWs), also known as lettered words (Liu, 1994) or code-mixing words (Nguyen and Cornips, 2016), are usually formed by Latin, Greek, Arabic alphabets in com- bination with Chinese characters, e.g. “X-光/X射 线”, X-ray. Although pure alphabets (e.g. “NBA”) used in Chinese context have also been regarded as MAWs in some previous work (Liu, 1994; Huang and Liu, 2017), they are more like switching-codes that retain the orthography and linguistic behav- iors of the original language, instead of showing typical Chinese lexical characteristics. It is note- worthy that MAWs shall be taken as a code-mixing phenomenon instead of code-switching as a MAW is still a Chinese word which is not switched into another language. Therefore, in this work, MAWS refer to the combined type which encodes both al- phabet(s) and Chinese character(s) in one word, such as “A型”, A-type, “PO主”, post owner, and “γ线”, Gamma Ray. It is linguistically-interesting and applicably- significant to investigate MAWs due to two main reasons. First, A MAW maintains part of the Chinese characteristics in morphology, phonology and orthography (e.g. “PK过", player killed, past tense). Meanwhile, it also demonstrates some prop- erties of the foreigner language (e.g. “维生素ing", supplementing Vitamin, progressive)), providing a unique lexical resource for studying morpho- phonological idiosyncrasies of code-mixing words. Second, MAWs serve as an indispensable part of people’s daily vocabulary, especially under the rapid development of social media communication. Yet, being out-liars of the Chinese lexicon, they can cause problems to existing word segmentation/new word extraction tools that are trained on traditional words (Chen and Liu, 1992; Xue and Shen, 2003). Consider the following example: E1: PO主 主 主也不知道链接被吞了 (The post owner didn’t know that the link has been hacked off) Seg: PO/主 主 主/也/不/知道/链接/被/吞/了 Golden Seg: PO主 主 主/也/不/知道/链接/被/吞/了 The sentence in E1 (example 1) is segmented using Stanford Parser …

Content truncated to 500 words