Automatic extraction of mobility activities in microblogs
Ricardo César de Sales Ferreira · 2012
The management and planning of public transport, urban planning and Marketing decision making are everyday activities and they are influenced by the mobility of people. So the problem we face is that to make a good urban planning, management of transport networks and even marketing strategies, we need to know the mobility patterns of people. For this, there are several sources such as surveys, census data, among others. The sources of information for urban areas are varied, but not complete. It is in this context that we want this project of dissertation to be a more complete source of mobility intentions, as much as it will present a more intuitive way than the existing sources. Previously sources of information that might reflect the patterns that we want to get were not abundant. Now, alongside the growth of information on the Internet, coupled with a greater use of social networks, there are several sources of knowledge capable of extracting this data set so that the information might have value to the user. Thus, with this dissertation, we intend to extract information from these sources, filter and treat it so that we may obtain an amount of knowledge about the activities of mobility. We intend to report the obtained results in the form of an intuitive and user friendly web interface. This document is itself an information source and it is the result of a study of related work in the area of the information extraction and current techniques. In this document we also describe the solution implemented. We evaluated a random sample of messages from Twitter to be classified as containing mobility activities or not and the results were a precision of 82.7 % and a recall of 62 %. It means that our priority was to improve precision than recall. In other words, we preferred to be correct in the messages that we classify as mobility rather than getting almost all the mobility messages, thus we improved our algorithm in that way. The greatest motivation for this dissertation is exactly the fact that it aims to contribute to a better planning and actual decisions for our country.