Search

Multi-task learning is becoming more popular and is being applied in a variety of applications. It improves the accuracy of prediction by simultaneously learning related tasks and saves cost through shared structures. In particular, the prediction of event type from social data is also an area where multi-task learning can be utilized. In this paper, we present a novel deep learning framework called S patial E vents P rediction (SEP) based on multi-task learning to predict the types of events that happen at a specific location from social data. The proposed model focuses on predicting the attribute types of an event, which is referred to as subtypes. Specifically, an event type-specific attention mechanism is introduced to extract the representations of social data and to identify their important components. The proposed attention mechanism is based on a two-level attention, which measures the importance of words and sentences to the subtypes of an event. We also propose a representation sharing method using semantic and spatial relationships between locations to alleviate the sparsity and incompleteness of data. The proposed representation sharing preserves the spatial heterogeneity between locations and significantly improves the accuracy of the overall framework. Experiments with real-world datasets confirm the effectiveness and efficiency of the proposed method.

Nowadays, question answering (QA) over complex questions is the most spotlighted research topic in diverse communities. Different from existing QA approaches for simple questions that require a single relation in a knowledge graph (KG) to find an answer set, QA for complex questions requires to process multiple KG relations. Complex questions often include some representations for constraints (ordinal, temporal, etc.) that restrict an answer set for the given question. Despite lots of efforts to process such questions, there are still limitations in processing constraints, multiple relations, and variables. To solve the issues, we propose a novel QA method that first decomposes an input question and then generates a correct query graph with fully complete semantics. In order to fill with lossy semantics caused by the decomposition task, the proposed method employs Bi-GRU based model. The model integrates individual compositional semantics of sub-query graphs matched to decomposed sub-questions. Experimental results show the best performance compared to the state-of-art on complex questions.