Tat Chua’s scientific contributions

In-game actions of real-time strategy (RTS) games are extremely useful in determining the players' strategies, ana-lyzing their behaviors and recommending ways to improve their play skills. Unfortunately, unstructured sequences of in-game actions are hardly informative enough for these analyses. The inconsistency we observed in human annotation of in-game data makes the analytical task even more challenging. In this paper, we propose an integrated system for in-game action segmentation and semantic label assignment based on a Conditional Random Fields (CRFs) model with essential features extracted from the in-game actions. Our experiments demonstrate that the accuracy of our solution can be as high as 98.9%.

The human language is inherently unstructured. Human authors never follow the strict rules of grammar because their human readers possess cognitive abilities to interpret the semantic meaning of unstructured human written text. Since unstructured text possess ambiguities and uncertain- ties, using probabilities to model human language is a nat- ural choice. The investigation of a widely used probabilistic model is the motivation of my survey here. This proba- bilistic model known as Latent Dirichlet Allocation (LDA), has seen widespread adoption in the fleld of information re- trieval. As prelude to the introduction of LDA, I shall re- view the Latent Semantic Analysis (LSA) and Probabilistic Latent Semantic Analysis (PLSA) for historical interests. Finally, I brie∞y discuss Hierarchical Latent Dirichlet Allo- cation.