Aggregation of fragmented location descriptions from the web for human way-finding

Abstract

This research proposes a novel methodology for aggregating various corpus of a certain location sourced from the blogosphere to generate a location description which can support human’s recognition and find the desired location. Location information is required to travel from one place to the other. People acquire location information from various sources in various forms. A typical one is an official address. This however is composed of street names and address numbers which makes it hard to intuitively find the location. Map is another source of location information which is in a form of a picture. However map does not directly give explicit information but by a cognitive translation process, we acquire the implicit location information. Typical implicit location information is the description directly given by a person. The location description created by people is landmark based description, because people tend to perceive a location by its relative landmarks. This location description created by one person alone may not be accurate or understandable due to its subjective characteristics. Therefore by sourcing a large number of various corpora of location descriptions from the blogosphere and by the aggregation scheme proposed, we generated location descriptions that are more accurate and objective for human way-finding. Four steps of generating location descriptions are defined: phrase extraction, phrase aggregation, phrase ranking and phrase selection and arrangement. The mechanism for each stages of the system was specified and finally evaluated for its effectiveness. The experimental results of this study showed that, in a comparison of the official address as well as the location descriptions of random bloggers, the recognition of the location description generated by the proposed scheme showed a higher level of accuracy and efficiency in time.