Authorship Attribution for Twitter in 140 Characters or Less

Abstract

Authorship attribution is a growing field, moving from beginnings in linguistics to recent advances in text mining. Through this change came an increase in the capability of authorship attribution methods both in their accuracy and the ability to consider more difficult problems. Research into authorship attribution in the 19 ℎ century considered it difficult to determine the authorship of a document of fewer than 1000 words. By the 1990s this values had decreased to less than 500 words and in the early 21 century it was considered possible to determine the authorship of a document in 250 words. The need for this ever decreasing limit is exemplified by the trend towards many shorter communications rather than fewer longer communications, such as the move from traditional multi-page handwritten letters to shorter, more focused emails. This trend has also been shown in online crime, where many attacks such as phishing or bullying are performed using very concise language. Cybercrime messages have long been hosted on Internet Relay Chats (IRCs) which have allowed members to hide behind screen names and connect anonymously. More recently, Twitter and other short message based web services have been used as a hosting ground for online crimes. This paper presents some evaluations of current techniques and identifies some new preprocessing methods that can be used to enable authorship to be determined at rates significantly better than chance for documents of 140 characters or less, a format popularised by the micro-blogging website Twitter 1 . We show that the SCAP methodology performs extremely well on twitter messages and even with restrictions on the types of information allowed, such as the recipient of directed messages, still perform significantly higher than chance. Further to this, we show that 120 tweets per user is an important threshold, at which point adding more tweets per user gives a small but non-significant increase in accuracy.

Convolutional Neural Networks for Authorship Attribution of Short Texts

“…We evaluate an LSTM trained on bigrams, since the LSTM produced better results on a small validation set. CHAR: Character and word n-grams have been the core of many AA systems (Stamatatos, 2009;Schwartz et al, 2013;Layton et al, 2010). We tested various n-gram combinations on the small validation set and our final system uses character 2,3,4-grams with logistic regression.…”

“…With the advent of social media, one can even argue that building systems that work with short texts equally, if not more important than long texts like books. This need is also reflected in the increasing interest in AA of small texts such as tweets and reviews in AA research community (Qian et al, 2015;Schwartz et al, 2013;Layton et al, 2010).…”

“…However, we were able to find research in AA using traditional as well as related approaches. Character and word n-grams have been used as the core of many authorship attribution systems (Stamatatos, 2009;Schwartz et al, 2013;Layton et al, 2010). Character and word n-grams help determine the author of a document by capturing the syntax and style of an author.…”

‘twazn me!!! ;(’ Automatic Authorship Analysis of Micro-Blogging Messages

“…More recently, it has been demonstrated that the authorship of twitter messages can be attributed with a certain degree of certainty [11]. Surprisingly, the authors concluded that authorship could be identified at 120 tweets per user, and that more messages would not improve accuracy significantly.…”

Linguistically Informed Tweet Categorization for Online Reputation Management

“…While the majority of research in the computational sciences on Twitter data has focused on issues such as topic detection (Cataldi et al, 2010), event detection, (Weng and Lee, 2011;Sakaki et al, 2010), sentiment analysis, (Kouloumpis et al, 2011), and other tasks based primarily on the topical and/or semantic content of tweets, there is a growing body of work which investigates more subtle forms of information represented in tweets, such as reputation and trustworthiness, (O'Donovan et al, 2012), authorship attribution (Layton et al, 2010;Bhargava et al, 2013) and Twitter spam detection, (Benevenuto et al, 2010).…”