Twitter Sports: Real Time Detection of Key Events from Sports Tweets
Journal Title: Transactions on Machine Learning and Artificial Intelligence - Year 2017, Vol 5, Issue 6
Abstract
tweets about them. Event detection in Twitter is the process of detecting an event which is an occurrence causing change in the volume of tweets that discuss the associated topic at a specific time and a location by Twitter users. Twitter has been extensively used to detect major social and physical events such as earthquakes, celebrity deaths, presidential elections, traffic jam and others. Real time event detection in Twitter is detecting real-life events from live tweets instantly as soon as the event has occurred. Real time event detection from Cricket sports using Twitter media is an interesting, yet a complex problem. Because, event detection algorithm needs live tweets streamed at real-time about the game and should detect events such as boundary and sixer, at near real-time within few seconds from their occurrences. In this paper, a novel real-time event detection approach is proposed for the Cricket sports domain. The proposed approach first computes the post rate of an adaptive window, which is the ratio between the volumes of tweets in the second half window and the volume of tweets in the first half. An event has occurred if the post rate is above the pre-defined threshold, otherwise the algorithm selects the next big window in an adaptive manner. The predefined threshold helps to filter out the small spikes in the streaming tweets volume. Once an event is detected in a time window along the tweet stream, the event represented inside the window is recognized using the event lexicon representing different events of a cricket game. The proposed real-time event detection algorithm is extensively evaluated on 2017 IPL T20 Cricket sports dataset using ROC and AUC evaluation measures. The experimental results on the performance of the proposed approach show that the adaptive sliding window detects sports events with over 80% true positives and around 15% false positive rates.
Authors and Affiliations
Jeyakumar Kannan, AR. Mohamed Shanavas, Sridhar Swaminathan
Keywords
Related Articles
- EP ID EP267761
- DOI 10.14738/tmlai.56.3729
- Views 79
- Downloads 0
Remote-controlled Laboratories of Experimental Physics: Measuring the Stiffness of a Spring
In modern ELearning, students and scientists will be able to access Web portals for scientific computer and data infrastructures, thus accessing large collections of data and digital objects using metadata, knowledge man...
Implementation of the Flexible "Private - Public" Cloud Solution based on OpenStack
Cloud computing is a model that facilitates access to and manipulation of resources on demand. It is a technology that is unique today to meet the needs and demands of customers by guaranteeing a high quality of service...
Face Spoofing and Counter-Spoofing: A Survey of State-of-the-art Algorithms
In the current scenario of biometric-based identity verification, a face is still being proved to be an essential physiological evidence for successful person identification without letting know the target. Nevertheless,...
Improved HMM for Cursive Arabic Handwriting Recognition System Using MLP Classifier
Recognizing unconstrained cursive Arabic handwritten text is a very challenging task the use of hybrid classification to take advantage of the strong modeling of Hidden Markov Models (HMM) and the large capacity of discr...
Stress Management in Primary Caregivers: A Health Challenge
Technology based advances in healthcare are leading to social changes that, in turn, will require innovative responses from health systems. Mainly due to these advances, life expectancy grows so, the aging of world popul...