TwiSeg NER-Tweet Segmentation Using Named Entity Recognition

Now a days Twitter has provided a way to collect and understand user's opinions about many private or public organizations. All these organizations are reported for the sake to create and monitor the targeted Twitter streams to understand user's views about the organization. Usually a user-defined selection criteria is used to filter and construct the Targeted Twitter stream. There must be an application to detect early crisis and response with such target stream, that require a require a good Named Entity Recognition (NER) system for Twitter, which is able to automatically discover emerging named entities that is potentially linked to the crisis. However, many applications suffer severely from short nature of tweets and noise. We present a framework called HybridSeg, which easily extracts and well preserves the linguistic meaning or context information by first splitting the tweets into meaningful segments. The optimal segmentation of a tweet is found after the sum of stickiness score of its candidate segment is maximized.This computed stickiness score considers the probability of segment whether belongs to global context(i.e., being a English phrase) or belongs to local context(i.e., being within a batch of tweets).The framework learns from both contexts.It also has the ability to learn from pseudo feedback. Also from the result of semantic analysis the proposed system provides with sentiment analysis.