NER for Opinion Summarization Over Twitter Dataset

NER is an important subtask in information extraction which suffers severely from the noisy nature and short length of tweets. Toward addressing these problems, Here NER system uses tweet segmentation, POS tagging, global and local context together to identify candidate named entities from the noisy Twitter stream and random walk model to identify final named entities. Also, the system generates a tweet summary for each identified named entity. The experimental results for tweet dataset with tweet segmentation, POS tagging, and different K values prove the effectiveness of this NER system.