Learning from Negative News by Substituting Negative Images with Word2vec – A new technique called negative image enhancement (NNE) has been proposed to exploit image attributes such as background, background color and foreground in a way that can increase the quality of a visual scene. However, only a limited amount of training data is available for the NNE approach. This paper proposes a novel approach based on the use of the image dimensionality score to enhance the quality of the image in a deep learning framework. We show that our proposed technique can effectively enhance the image in the same way as the image dimensionality score would enhance. The evaluation on several popular image enhancement benchmarks shows that our proposed method significantly improves performance compared to other similar approaches.
We propose an end-to-end Semantic Web search framework designed to bridge the gap between machine learning and the web, based on a novel architecture for the process of searching for semantic entities. The approach takes a semantic ontology (MOT) as its semantic domain and a deep learning approach (DLA), as input, respectively. This model has two major advantages: (i) it can simultaneously capture semantic information and build a search engine with the same semantic capabilities, but also it is able to build one-to-many queries which can be easily extended to large, heterogeneous search environments without the need of query-specific knowledge. (ii) The framework is able to handle complex search environments using simple query-specific queries, and can provide an effective user interface with user-friendly user interface highlighting. We demonstrate the performance of the framework on the Web-15 benchmark, where we outperform the current state-of-the-art on both the task of query-based search and search related tasks.
Deep Learning for Human Action Detection: Dataset and Experiments
A Discriminative Model for Relation Discovery
Learning from Negative News by Substituting Negative Images with Word2vec
Lifelong Learning for Answerability Education
Can natural language processing be extended to the offline domain?We propose an end-to-end Semantic Web search framework designed to bridge the gap between machine learning and the web, based on a novel architecture for the process of searching for semantic entities. The approach takes a semantic ontology (MOT) as its semantic domain and a deep learning approach (DLA), as input, respectively. This model has two major advantages: (i) it can simultaneously capture semantic information and build a search engine with the same semantic capabilities, but also it is able to build one-to-many queries which can be easily extended to large, heterogeneous search environments without the need of query-specific knowledge. (ii) The framework is able to handle complex search environments using simple query-specific queries, and can provide an effective user interface with user-friendly user interface highlighting. We demonstrate the performance of the framework on the Web-15 benchmark, where we outperform the current state-of-the-art on both the task of query-based search and search related tasks.