Detecting Cyberbullying Entries on Informal School Websites Based on Category Relevance Maximization

International Joint Conference on Natural Language Processing, pages 579–586, Nagoya, Japan, 14-18 October 2013. Detecting Cyberbullying Entries on Informal School Websites Based on Category Relevance Maximization Taisei Nitta † Fumito Masui † Michal Ptaszynski † Yasutomo Kimura § Rafal Rzepka ‡ Kenji Araki ‡ † Department of Computer Science, Kitami Institute of Technology nitta@ialab.cs.kitami-it.ac.jp, {f-masui,ptaszynski}@cs.kitami-it.ac.jp § Department of Information and Management Science, Otaru University of Commerce kimura@res.otaru-uc.ac.jp ‡ Graduate School of Information Science and Technology, Hokkaido University {kabura,araki}@media.eng.hokudai.ac.jp Abstract We propose a novel method to detect cy- berbullying entries on the Internet. “Cy- berbullying” is defined as humiliating and slandering behavior towards other people through Internet services, such as BBS, Twitter or e-mails. In Japan members of Parent-Teacher Association (PTA) per- form manual Web site monitoring called “net-patrol” to stop such activities. Unfor- tunately, reading through the whole Web manually is an uphill task. We propose a method of automatic detection of cyber- bullying entries. In the proposed method we first use seed words from three cat- egories to calculate semantic orientation score PMI-IR and then maximize the rele- vance of categories. In the experiment we checked the cases where the test data con- tains 50% (laboratory condition) and 12% (real world condition) of cyberbullying en- tries. In both cases the proposed method outperformed baseline settings. 1 Introduction “Cyberbullying” is a new form of bullying. It is carried out on the Internet instead of classrooms. It takes form of hate messages sent through e-mails, electronic Bulletin Board System (BBS), etc., with the use of personal computers or mobile phones. Recently it has become a serious social problem in many countries, one of which is Japan (MEXT, 2008). Examples of cyberbullying that actually happened, include ridiculing student personality, body type, or appearance on informal school BBS, slandering students and insinuating they had per- formed deviate sexual intercourses. Some cases of cyberbullying lead the students who were bul- lied to assault or kill themselves or the student who wrote the bullying entry on the BBS. To deal with the problem members of Parent- Teacher Association (PTA)1 perform Web site monitoring activities called “net-patrol”. When a harmful entry is detected the net-patrol mem- ber who found it sends a request to remove the entry to the Internet provider or Web site admin- istrator. Some of the actual examples of harmful entries which were requested for deletion are rep- resented in Table 1 (names, phone numbers and other personal information was changed). Unfortunately, net-patrol has been carried out mostly manually. It takes much time and effort to find harmful entries (entries that contain harm- ful information and expressions) in a large amount of contents appearing on countless number of bul- letin board pages. Moreover, the task comes with a great psychological burden on mental health to the net-patrol members. To solve the above prob- lem and decrease the burden of net-patrol mem- bers, Matsuba et al. (2011) proposed a method to detect harmful entries automatically. In their method they extended the method of rel- …

Content truncated to 500 words