Warning: Parameter 1 to Language::getMagic() expected to be a reference, value given in /usr/share/mediawiki/includes/StubObject.php on line 58
Information extraction for disaster management – GRK-Wiki

Information extraction for disaster management

Aus GRK-Wiki

Wechseln zu: Navigation, Suche

Problem Statement

After a disaster like an earthquake, there is a huge demand for information by decision makers. More and more, such information can be found in textual form on the Internet, both in conventional sources like newspapers and in modern media such as blogs and newsgroups or social networks like Twitter and Flickr. These sources offer among the most detail information available, but their manual analysis is a time-consuming and therefore costly task. Information extraction (IE) is an area of research dealing with the problem of automatically extracting structured information from unstructured text. The methods used are mostly pattern matching, natural language processing, and machine learning. Also the task described before can be seen as an IE problem. This leads to the following research questions: (I) Which IE methods are the most appropriate ones to tackle textual messages in the context of (natural) disasters? (II) What quality of results can be expected? (III) How can the extracted information be combined with other resources, like geospatial databases or maps, to increase the usefulness of the data?


Approach

The field of natural language processing offers several approaches to enrich character data (text) by morphological, syntactical or semantical annotations like part of speech, phrase structures, dependencies or word senses. Those annotations can be utilized by machine learning techniques like kernel machines or pattern matching to reveal some of the demanded semantic information. A specific problem in the tackled domain is the predominant need to address n-ary relationships. Furthermore, addressing web content carries the danger that texts are ungrammatical, which might break NLP methods.

Persönliche Werkzeuge
Sprache