Text Mining from Free Unstructured Text: An Experiment of Time Series Retrieval for Volcano Monitoring

Volcanic activity may influence climate parameters and impact people safety, hence monitoring its characteristic indicators their temporal evolution is crucial. Several databases, communications literature providing data, information updates on active volcanoes worldwide are available, will likely increase in the future. Consequently, extraction text mining techniques aiming to efficiently analyze such databases gather data of interest a specific volcano can play an important role this applied science field. This work presents natural language processing (NLP) system that we developed extract geochemical geophysical from free unstructured included reports operational bulletins issued by volcanological observatories HTML, PDF MS Word formats. The NLP enables relevant gas (e.g., SO2 CO2 flux) text, was tested series 2839 daily weekly published online between 2015 2021 for Stromboli (Italy). experiment shows proves capable time set user-defined be later analyzed interpreted specialists relation with other geospatial data. potentially tuned target databases.