Analysis of human-wildlife interactions from social media and news sources: A hybrid neural network approach

从社交媒体和新闻来源分析人类与野生动物的相互作用：混合神经网络方法

• 批准号: 2807448
• 金额: --
• 依托单位: University of Manchester
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2022
• 资助国家: 英国
• 起止时间: 2022 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2807448
• 关键词: Analysis; human; wildlife; interactions; social

This project aims to create a hybrid neural network model consisting of Convolution Neural Network (CNN) to extract character level features with Bisectional Long-Short Term Memory (BiLSTM) to analyse and extract rescue and response information from real time Twitter data and news sources.The project will answer the following questions:1. How can the semantics of Tweets and News sources be analysed for sentiment without loss of context?2. How can features such as location be extracted from raw data? How would sequence modelling connect and preserve context of the extracted feature of interest?3. How could the proposed model "remember" past knowledge and connect it with the post-fire response stage of physical wildfire models? 4. How can the insights help build wildfire response and mitigation policies in the UK and India?

该项目旨在创建一个由卷积神经网络（CNN）组成的混合神经网络模型，以提取字符级特征，并使用二分长短期记忆（BiLSTM）从真实的时间Twitter数据和新闻源中分析和提取救援和响应信息。如何在不丢失上下文的情况下分析推文和新闻源的语义以获得情感？2.如何从原始数据中提取位置等特征？序列建模如何连接和保留提取的感兴趣特征的上下文？3.该模型如何“记住”过去的知识，并将其与物理野火模型的火灾后响应阶段联系起来？4.这些见解如何帮助英国和印度制定野火应对和缓解政策？