DataBridge: Bridging Data Using Sociometric Approaches

The fact that some scientific data are collected by scientists from different research domains makes them seemingly isolated and increases the difficulty of usage. Such data is also referred to as the long tail of science data. The DataBridge system is a scientific collaboration environment that realizes the potential of such data by implementing algorithms and tools to more easily enable data discoverability and reuse. The system has two computational modules known as the sociometric network analysis (SNA) module and the social computing (SC) module. The SNA module consists of sociometric network analysis algorithms for similarity network generation as well as similarity network reduction. The SC module applies community detection algorithms to detect malicious users. It also utilizes a sentiment analysis methodology to assist user opinion analysis, which can be used to further improve the algorithms in the SNA module. Specifically, for the SNA module, we propose a semantic similarity measure for the evaluation of existing semantic similarity measures; we also present a novel semantic similarity measure based on a recurrent neural network language model; a network reduction method is present and applied in order to simplify similarity networks. For the SC module, our contributions consist of a community detection framework for detecting malicious users, where the framework utilizes information reliability as the measure of a user's behavior, and a sentiment analysis methodology, where it suggests a full process for how to analyze user opinions using machine learning approaches. We also successfully perform extensive experimental evaluations for each of the proposed topics.