2019/09/01 by Marco Del Tredici, Diego Marcheggiani, Del Tredici, Marco +5
Computer Science · #Topic Modeling #Advanced Graph Neural Networks #Sentiment Analysis and Opinion Mining
paper · pdf · doi:10.48550/arxiv.1909.00412
Information about individuals can help to better understand what they say,\nparticularly in social media where texts are short. Current approaches to\nmodelling social media users pay attention to their social connections, but\nexploit this information in a static way, treating all connections uniformly.\nThis ignores the fact, well known in sociolinguistics, that an individual may\nbe part of several communities which are not equally relevant in all\ncommunicative situations. We present a model based on Graph Attention Networks\nthat captures this observation. It dynamically explores the social graph of a\nuser, computes a user representation given the most relevant connections for a\ntarget task, and combines it with linguistic information to make a prediction.\nWe apply our model to three different tasks, evaluate it against alternative\nmodels, and analyse the results extensively, showing that it significantly\noutperforms other current methods.\n