Arabic to English Person Name Transliteration using Twitter

Arabic to English Person Name Transliteration using Twitter Hamdy Mubarak, Ahmed Abdelali Qatar Computing Research Institute Hamad Bin Khalifa University (HBKU), Doha, Qatar {hmubarak, aabdelali}@qf.org.qa Abstract Social media outlets are providing new opportunities for harvesting valuable resources. We present a novel approach for mining data from Twitter for the purpose of building transliteration resources and systems. Such resources are crucial in translation and retrieval tasks. We demonstrate the benefits of the approach on Arabic to English transliteration. The contribution of this approach includes the size of data that can be collected and exploited within the span of a limited time; the approach is very generic and can be adopted to other languages and the ability of the approach to cope with new transliteration phenomena and trends. A statistical transliteration system built using this data improved a comparable system built from Wikipedia wikilinks data. Keywords: Transliteration, Named Entities, Social Media, Tweet Normalization, Arabic Language Variations 1. Introduction With the emergence of social media outlets, millions of users exchange messages daily. This rapid expansion raises new challenges related to retrieval and extraction in a multilingual scope. Named Entities processing has been recognized as a key technique that supports a number of Natural Language Processing fields (Callan and Mitamura, 2002) and (Khalid et al., 2008). Us- ing traditional approaches for building transliteration resources (Kirschenbaum and Wintner, 2010; H´alek et al., 2011) or mining them from text and news (Darwish et al., 2012; Kumaran et al., 2010; Sajjad et al., 2011) might not keep the pace with rapid expansion of infor- mation form such outlets. The social media outlets are providing large volume, high-value, content that is being sought by researchers, both in business and academia. Opinion mining (Lukasik et al., 2015; Manoochehr et al., 2013; Agarwal et al., 2011), customer relation, eBusi- ness, eHealth (Paul and Mark, 2011; Luis et al., 2011) are examples for disciplines that are exploiting these resources. The amount of data generated from the tweets only surpasses 500 millions tweets per day1, as such, it presents a unprecedented type of versatile resource that can be uti- lized namely for transliteration. Unlike similar resources, Twitter data includes explicit data about user, location, language, social network,..etc. In our paper, we present results of experiments for har- nessing large number of tweeps 2 information to build a transliteration module that can be used to support transla- tion as well as cross-language information retrieval. The advantage of using tweets versus other methods is the ac- curacy as well as the freshness. While linguistic resources such as Encyclopedia, Onomasticons might require time to maintain and update. Social media are becoming a faster way to get large amount of information. The occurrence 1See http://www.internetlivestats.com/twitter-statistics/ 2Tweep: A person who uses the Twitter online message service to send and receive tweets. frequency of a given item reflect well the accuracy and its standard use. For our case-study language “Arabic”, we were able to collect over 880,000 unique Arabic users with their transliteration to English in a period …

Content truncated to 500 words