Semantic Annotation of Social Media Data Sample Clauses

Semantic Annotation of Social Media Data. Current social media annotation techniques use event specific patterns based on event types (Xxxxxxxx et al., 2002)(Xxxxx et al., 2008). The text is expected to have some structure (e.g., news documents). Such a technique does not scale for city events from twitter text due to the informal nature of tweets. Further, the aggregation is done at a cluster level, which is too coarse grained for city related events. Important city events may be reported by few citizens given the wide scope of topics on twitter (Xxxx et al., 2010). In such situations, clustering based techniques (Xxxxxxxx et al., 2006) may fail to separate minority tweets related to the city infrastructure. Twitter messages may be noisy and convoluted with little context in which the message was generated. Such a characteristic of tweets challenges a dictionary-based approach for spotting location and event terms. For example, Figure 3 shows one of many instances where a single event term ‘accident’ is used in different contexts.