Abstract In the last few years, there is an increase of IoT applications in several areas including smart homes, smart cites and so on which ensure entity-based data streams and work on real-life. While the utilization of the resources is assured, these applications need more notification services to users. Some of these needs can be given by the Publish or Subscribe... In the last few years, there is an increase of IoT applications in several areas including smart homes, smart cites and so on which ensure entity-based data streams and work on real-life. While the utilization of the resources is assured, these applications need more notification services to users. Some of these needs can be given by the Publish or Subscribe systems; however, the alternative smart systems can reinforce such systems. They can able to increase the reliability or usability and overcome redundancy awareness. In this research we intend to create a boundary for entity centric Publish or Subscribe systems and work for data fusion and top-k diverse ranking. We experimented our system and found that the top-k fused diverse summary yielded better results than separate notifications and without top-k filtering. The top-k fused diverse summarisation scores well in the testing and an acceptable F-score is arrived. The inferences came at the expense of higher latency and there is some equation lies between latency, the number of forwarded messages, and expressiveness. Read More Read Less

Abstract Extraction of data as entities from information descriptions is the main task in the entity analysis. Conventional techniques include clustering and graph whereas the current exercise tries to integrate deep learning in the model. The proposed model is a neural network model where we basically create a supervised mechanism. We measure and compute attention weights for facts. We then rank... Extraction of data as entities from information descriptions is the main task in the entity analysis. Conventional techniques include clustering and graph whereas the current exercise tries to integrate deep learning in the model. The proposed model is a neural network model where we basically create a supervised mechanism. We measure and compute attention weights for facts. We then rank factual date for generate summaries. We work with methods to fix complex learning problems. While testing with extensive applications, the proposed technique increases the quality of entity summaries. This is arrived from the results of both F-measure and MAP compared with some state-of-the-art methods. We have given the description with source codes at https://github.com/WeiDongjunGabriel/ESA1. Read More Read Less

Abstract We in this work have worked for data retrieval from Chinese historiography. The main issue is the low resource of the language: deep learning requires large amounts of annotated data and becomes impracticable when such data is not available. We used the subject experts to curate a set of person entities and their profile attributes and relations from two documents. We... We in this work have worked for data retrieval from Chinese historiography. The main issue is the low resource of the language: deep learning requires large amounts of annotated data and becomes impracticable when such data is not available. We used the subject experts to curate a set of person entities and their profile attributes and relations from two documents. We introduce a pattern-based bootstrapping approach to extract the information with a very small number (i.e., 1 or 2) of seed patterns. The testing results show the effectiveness as well as the limitations of the iterative method. Read More Read Less