2017/09/13 by Tu Ngoc Nguyen, Nguyen, Tu, Cheng Li +4 · 1 citation
Social Sciences · Physics and Astronomy · Computer Science · #Misinformation and Its Impacts #Complex Network Analysis Techniques #Advanced Text Analysis Techniques
paper · pdf · doi:10.48550/arxiv.1709.04402
Recently a lot of progress has been made in rumor modeling and rumor\ndetection for micro-blogging streams. However, existing automated methods do\nnot perform very well for early rumor detection, which is crucial in many\nsettings, e.g., in crisis situations. One reason for this is that aggregated\nrumor features such as propagation features, which work well on the long run,\nare - due to their accumulating characteristic - not very helpful in the early\nphase of a rumor. In this work, we present an approach for early rumor\ndetection, which leverages Convolutional Neural Networks for learning the\nhidden representations of individual rumor-related tweets to gain insights on\nthe credibility of each tweets. We then aggregate the predictions from the very\nbeginning of a rumor to obtain the overall event credits (so-called wisdom),\nand finally combine it with a time series based rumor classification model. Our\nextensive experiments show a clearly improved classification performance within\nthe critical very first hours of a rumor. For a better understanding, we also\nconduct an extensive feature evaluation that emphasized on the early stage and\nshows that the low-level credibility has best predictability at all phases of\nthe rumor lifetime.\n