Named Entity Recognition Supporting Serious Games Development In Stack Overflow Social Content

Q&A sites for developers (like Stack Overflow) provide various types of game-specific social contents that can support the development of effective serious games - games that can really train, educate, and motivate players. These social contents include valuable information for effective serious games development, such as design principles and best practices, game-specific algorithms, game engines, API documentation and game libraries. However, existing techniques and tools for analysing social contents are mainly focusing on recognizing person, location, and organization, and thus are not designed to support the recent advance of entity-centric search systems, such as direct answers and knowledge-graph for serious games domain. In this research, we study the problem of NER to support the development of serious games. We address the challenges of recognizing game-specific named entities in social contents and develop a machine-learning based model that can recognize a broad category of named entities that game developers really care about. We conduct systematic experiments to evaluate our machine learning-based NER against a well-designed rule-based baseline system and to study the effectiveness of various NER techniques and features against the unique challenges of game-specific social contents.