Social text warehousing with topic graphs for fire damage analysis

A social network service is a platform where millions of users share their opinions in pre-formed social relations. Many SNS users generate huge amounts of textual data on the social networks in real time and the usergenerated texts are spread out rapidly. However, due to the limitation of short sentences on SNS, it is hard to clearly understand whether a given sentence is related to a particular subject area (e.g., fire damage) to be analyzed. In this paper, we introduce a new way of building a social text warehouse for fire damage analysis. In terms of fire damage analysis, it is highly significant to integrate SNS textual data and existing structured data. To extract keywords or topics on SNS clip related to the fire damage, we selected high-quality documents on Wikipedia by using the search engine called Elasticsearch and built the topic DAG with regards to the fire damage analysis using corpus-dependent topic hierarchy graph on Wikipedia.