User participation in building language repository: the case of Google Translate

2017 1st International Conference on Engineering and Applied Technology

Thamrin, H.^1 ; Ariyanto, G.^1 ; Pamungkas, E.W.^1 ; Sulistyono, Y.^2

Department of Informatics, Universitas Muhammadiyah Surakarta, Jl. A Yani Pabelan, Kartasura, Sukoharjo

57162, Indonesia^1

Department of Education of Bahasa Indonesia, Universitas Muhammadiyah Surakarta, Jl. A Yani Pabelan, Kartasura, Sukoharjo

57162, Indonesia^2

关键词: Data repositories;    Google translate;    In-buildings;    Language resources;    Lexicon-based;    Sentiment scores;    Translation algorithms;    User participation;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/403/1/012072/pdf
DOI  :  10.1088/1757-899X/403/1/012072

来源: IOP

【 摘 要 】

A language repository is a place that stores language resources, such as dictionary or corpus. Accurate and large size repository is important to improve the performance of natural language processing applications, especially those that employs repository-based algorithms. This research tries to confirm the possibility of building language repository in Bahasa Indonesia involving user participation. We survey 191 people about their experience and opinion in using Google Translate, at the time they first used the application several years ago and recently. Opinions were then analyzed for sentiment score using lexicon-based sentiment analysis. The shift of sentiment was calculated and confirmed using t-test. There is a change in opinion score, from 0.3174 at first time use of the application to 0.4980 at recent use. The improvement can be attributed to the introduction of better translation algorithm and the improvement of data repository (or dictionary) by crowdsourcing. We conclude that it is possible to involve users to build language repository for Bahasa Indonesia.