CREES automatically annotates short texts. It identifies if a text is about a crisis, crisis-types and information-types.
The Crisis Event Extraction Service (CREES) provides multiple functions for annotating short text documents for helping to classify large amount of short document (e.g. tweets) by 1) identifying if a document is related to a crisis event (e.g., fire, earthquake); 2) the type of event discussed and, 3) the type of information present in a document.
The CREES add-on provides three different customs functions that can be applied on textual columns or individual cells for annotating the contained text. Each cell should contain a small document similar in length to tweets.
The CREES add-on makes available the following functions:
- CREES_RELATED: Identifies if a short text document discusses a crisis event (returns one of the following labels: non-related, related).
- CREES_EVENT: Identifies the type of event discussed within a short textual document (returns one of the following labels: bombings, collapse, crash, derailment, earthquake, explosion, fire, floods, haze, meteorite, none, shootings, typhoon, wildfire).
- CREES_INFO: Identifies the type of information discussed within a short textual crisis-related document (returns one of the following labels: affected_individuals, caution_and_advice, donations_and_volunteering, infrastructure_and_utilities, not_applicable, not_labeled, other_useful_information, sympathy_and_support).
More information about each function can be obtained in the built-in documentation. You can access it by typing its the function name in a spreadsheet cell.
More information about CREES can be found on the CREES website (https://evhart.github.io/crees/) or in the following publication:
On Semantics and Deep Learning for Event Detection in Crisis Situations. Burel, Grégoire; Saif, Hassan; Fernandez, Miriam and Alani, Harith (2017). On Semantics and Deep Learning for Event Detection in Crisis Situations. In: Workshop on Semantic Deep Learning (SemDeep), at ESWC 2017, 29 May 2017, Portoroz, Slovenia.
This work has received support from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 687847 (COMRADES).