high performance computing on graphics processing units: hgpu.org

The goal of a Knowledge Base-supported Question Answering (KB-supported QA) system is to answer a query natural language by obtaining the answer from a knowledge database, which stores knowledge in the form of (entity, relation, value) triples. QA systems understand questions by extracting entity and relation pairs. This thesis aims at recognizing the relation candidates inside a question. We define a multi-label classification problem for this challenging task. Based on the word2vec representation of words, we propose two convolutional neural networks (CNNs) to solve the multi-label classification problem, namely Parallel CNN and Deep CNN. The Parallel CNN contains four parallel convolutional layers while Deep CNN contains two serial convolutional layers. The convolutional layers of both the models capture local semantic features. A max over time pooling layer is placed on the top of the last convolutional layer to select global semantic features. Fully connected layers with dropout are used to summarize the features. Our experiments show that these two models outperform the traditional Support Vector Classification (SVC)-based method by a large margin. Furthermore, we observe that Deep CNN has better performance than Parallel CNN, indicating that the deep structure enables much stronger semantic learning capacity than the wide but shallow network.