整体概括: 这篇论文研究的基于传播路径的假新闻探测问题, 研究的场景是社交媒体, 以Twitter为例. 本文提出模型, 通过新闻的传播路径(最源头的推文文本信息, 转推者(retweet)的用户信息等)来做二分类问题. 论文提出的模型叫做Graph-aware Co-Attention Networks. 由此可以看出, 模型中用到了图(graph)相关的表征模型和co-attention的机制. 笔者的深度学习知识储备浅薄, 看到文章标题的第一反应是, 模型的可解释性(interpretability) 来自co-attention机制, 而图表征的技术手段则提供了一个新的表征维度, 原文作者一定是因为其给模型带来的增益较大, 才将其作为文题. 传统而言, 假新闻的表征手段关注于表征文本(源文本, 转发文本, 评论等等), 而本文表征的是传播路径. 一起来看看原文作者是怎么做的吧.
This tutorial is modified based on my repository customize-financial-plotly-plot and focuses on further beautifying and adding more functionalities into the plotly plot created in the last tutorial for my wife. We know that adjusting and choosing among several parameters’ combinations is a tedious and time-consuming work. So I hope the people who read this tutorial will not struggle to button-or-layout-related problems when plotting k-lines.
EXCLUSIVE! This tutorial was originally created for my beautiful and cute wife learning about Python Data Visualization in some financial cases. Well, this tutorial aims to finish THREE popular tasks concerning data analysis, data visualization and combining it into html file, aka. for the purpose of using the analysis elsewhere other than local computer.