Classification of Issues in the Public Space Using Their Textual Description and Geo-Location[go to overview]
Vandalism, environmental pollution or other damage to community property are among the daily problems, local communities have to deal with. In some countries governments use their residents to take on these problems. After spotting some- thing that needs to be fixed, citizens can use a free smartphone app to take a picture, select a location and an issue type and send the issue to the city council. This way issues from different categories like e.g. litter on the street, broken street lights or potholes can be sent directly to the department which is in charge of the respective issue category. Past experience with these apps however has shown that often, users are unwilling to provide the correct category when reporting an issue. This led to additional work load for civil servants who had to manually forward the reports to the correct department. Therefore, in this thesis it is examined in which way issue reports can best be automatically assigned to the correct category. The presented approach makes use of the textual description, the creation time and the geo-spatial context of the reported issues and is evaluated on two different data sets.
17.09.15 - 10:15