Climate Event Detection Algorithm Based on Climate Category Word Embedding
Hengyi Wang, Zhendong Niu · 2018
Detecting climate events efficiently and accurately is important in traffic warning, disaster warning, and disease prevention. Given that ordinary event detection algorithms are ignored in climate domain's knowledge, the results of climate domain's event detection are not satisfactory. This paper proposes a climate event detection algorithm based on climate category word embedding(CEDCWE). This method combines the climate category word embedding and typical factors of climate events as a document representation model to express detailed information of climate documents and detect climate events efficiently and accurately. Compared with other methods, the CEDCWE algorithm can generate better climate document representations and climate event detection results. In the experiments, we acquire the datasets by a web crawler and evaluate our CEDCWE on real-world climate event detection tasks. Experimental results show that our CEDCWE is effective in climate document representation and outperforms typical methods.