<![CDATA[Twitter is on of many social medias where the user often tell about events that are happenings around them, ranging from insignificant to important things. Twitter's qualities where the user could make a tweet anywhere in a brief time frame, make it feasible for critical information to appear before the media even report it. However, it is difficult to comprehend what relevant events are occuring in a specific region because of the sheer size of scale and diverse sort of tweets. Accordingly, there is a need of a framework that could do pertinent event detection and give a summary about that event. In light of the reason expressed over, this research center around text summarization for event detection of Indonesian Twitter archive utilizing Affinity Propagation. Through the process of clustering, the resulting clusters become the representation of events occuring in a specific place and time period. Two kinds of data are used for assessment, first is themathic which has spesific kind of event happening in the time frame of the tweet and second is generic where the tweet are taken from an arbitary time frame. In order to get the best resulting cluster, parametesr of Affinity Propagation are evaluated reuslting in preference of quartile 3 dan minimum, damping factor of 0,3 and 0,5, changed limit of 1 and 2, iteration maximum of 250 as the best parameters for the thematic and generic data. The result of tweet summary from the clustering process are then compared with a specialist's summary to be evaluated by ROGUE-N method, scoring 0,459 and 0,4009 respectively on two kinds of data.]]>
Copyrights © 2019
