Text Analysis for Understanding Gamer Social Behavior

电子游戏已经成为每个人生活的一部分，特别是在北美，超过95%的儿童和青少年经常玩游戏。随着所有游戏平台(从手机到专用游戏机)处理能力的不断增强，视频游戏变得越来越多样化、复杂性、创造性和社会性。现代游戏固有的社交方面一直是青年心理学家研究的主题，并发现它有助于许多积极的行为以及儿童和青少年的发展。然而，在线社区也有可能允许有毒的社会行为不被发现，并可能导致自我伤害或其他形式的不良行为。毒性的例子包括骚扰、欺凌、污名化和歧视。BioWare是一家屡获殊荣的加拿大视频游戏开发工作室，致力于为所有玩家提供安全的在线社区。为此，与艾伯塔大学的这项Engage赠款旨在开发和评估检测玩家社区中的有毒行为的算法，这些行为以短信或语音对话的形式表达。为此，我们将通过这个项目收集玩家社交互动产生的相关文本语料库，定义有效的毒性度量标准，并使用自然语言处理和机器学习中的最新方法来评估它们对检测有毒行为的适用性。我们预计，该项目将为语言处理的新方法带来新的想法，并就如何让参与者更好地理解他们的行为并推动他们远离有毒和有害的行为进行长期的后续工作。

Video games have become part of everyone's lives, especially in North America, where more than 95% of children and adolescents engage in gaming regularly. With a continuous increase in processing power of all gaming platforms, from phones to dedicated consoles, video games have become increasingly varied, complex, creative, and social. The inherently social aspect of modern gaming has been the subject of study by youth psychologists and found to contribute to many positive behaviors as well as to the development of children and adolescents. However, online communities also have the potential for allowing toxic social behavior going undetected, and potentially contributing to self harm or other forms of undesirable behavior. Examples of toxicity include harassment, bullying, stigmatization, and discrimination. BioWare, an award winning Canadian video game development studio, is committed to offering safe online communities for all its players. To that end, this engage grant with the University of Alberta seeks to develop and evaluate algorithms for detecting toxic behavior in a community of players, expressed as text messages or voice conversations. To that end, through this project we will collect relevant corpora of text arising from social interactions of players, define effective metrics of toxicity, and use state-of-the-art methods in Natural Language Processing and Machine Learning to assess their suitability for detecting toxic behavior. We anticipate the project will lead to novel ideas for new methods in language processing as well as a longer-term follow up work on ways to engage the players to better understand their actions and nudge them away from toxic and harmful behavior.