Online named entity recognition method for microtexts in social networking services: A case study of twitter

Named entity recognition (NER) methods have been regarded as an efficient strategy to extract relevant entities for answering a given query. The aim of this work is to exploit the conventional NER methods for analyzing a large set of microtexts of which lengths are short. Particularly, the microtexts are streaming on online social media, e.g., Twitter. To do so, this paper proposes three properties of contextual association among the microtexts to discover contextual clusters of the microtexts, which can be expected to improve the performance of NER tasks. As a case study, we have applied the proposed NER system to Twitter. Experimental results demonstrate the feasibility of the proposed method (around 90.3% of precision) for extracting relevant information in online social network applications.

► It is difficult for texts in the social media, e.g., Twitter, to conduct named entity recognition (NER) process.
► We propose an efficient contextual association discovery mechanism to support the NER process.
► Temporal association has shown the best performance out of three types of heuristic associations.