Mapping of aliases to points of interest

Abstract

In 2006, the social networking messaging service was created and allowed users to publish short messages of 140 characters. Through these short messages, also known as tweets, it allows users to publish their thoughts, opinions, daily activities and other valuable information. As users sometimes explicitly and implicitly reveal their location through the tweets, we can extract valuable information such as a user’s current location or locations they have been at a very fine granularity. Being able to recognise such location mentions and then mapping it to a well-defined location is a huge challenge. Tweets are often short, containing only 12-16 words on average, ungrammatical and informal in nature and locations are often mentioned with acronyms, incomplete names and slang terms. In this solution, it will first identify potential location aliases (variations of location entity mention) found within a Tweet itself using grammatical features through a trained classifier. In order to overcome the lack of context in Tweet, I have utilised the Google search engine to enrich the Tweet itself and post-process the potential aliases identified. Moreover, the solution will be able to map a specific location entity (which will be referred to as Point of Interest[POI]) with an alias. Finally, a list of POIs and aliases for each tweet will be outputted. In the evaluation section, you will be able to find the performance of the proposed solution.