Detection of Municipal Heat Plan Documents Using Semantic Recognition Methods
摘要
Municipalities in Germany are required by law to prepare a report on municipal heat planning by June 2028 at the latest, depending on their population. Due to the federal structure of Germany in some federal states there exist different regulations and due dates. Most of the municipalities with already completed municipal heat plans have published them on their respective websites. Neither the heat plans themselves nor the location of publishing follow generic templates and are therefore presented in different formats, lengths and places. In order to gain an overview of these heat plans in an effort to coordinate heat planning across municipalities in Germany, a data set referring to the available heat plans shall be created and regularly updated. The first step is to use an internet search engine and a web crawler to identify candidate documents for a municipality’s heat plan. In a second step, the results are checked for plausibility, i.e. it is checked whether the candidate documents are actually municipal heat plans and whether they are assigned to the correct municipality. The third step involves semantic enrichment by a process that includes the normalization of time expressions to extract important information from the municipal heat plan documents that relate to time sequences, as well as the extraction of geographical units to link the document to the correct municipality. A Web interface provides access to the detected municipal heat plans for evaluation and research purposes.