Trend Identification in Microblogs

Abstract

In the contemporary era, Society and Media are indispensable to each other and hence Social Media's tremendous popularity attracts the special scientific research & analytical study. The 'exponential growth' in the usage of a smartphone has elevated standard of living of multitude as part of their natural habits. Social media has transformed the way of daily communication. The latest list of tunnels of communication grades "Social Media" as an ineluctable tunnel considering its huge importance through which majority earth population share their beliefs, feelings, experiences, liking - disliking, dissatisfaction, hatred, love, feedbacks, and infinite human traits can be added here. During last decade, this medium has established itself as a strong medium of communication which represents the true voice of society. Before a decade, people used to spare a lot of time and interest for reading but nowadays competitive world has encompassed every one’s life. So no one has the abundance of time for all these. Here comes the beautiful and effective concept of Microblogging. Fulfilling all demands of contemporary social media, it has emerged as the most popular application. The user can express their thoughts in nutshell through specific character limits (Until September 2017, Twitter had 140-character limit for Tweets, now it is increased to 280 characters). The lesser the length of the script, the more complicated it is for anyone to track the meaning of the script. There is no doubt about the usage of Microblogs, as anyone can get a satellite- cum-street view about, “What people are thinking?” “what people are discussing?”, “what is happening in society?”. Answer or hint to above questions (through analysis of Microblog posts), helps to identify problems of society, hot topics in discussion, honest feedbacks of any product / service thus making it very much useful for business analysts, economist, researchers, political parties, government bodies to ‘catch’ the taste of society within short span. Here, the extract that we derive from the posts can be labeled as “Trend in Microblogs”. By analyzing Tweet behaviors, we can notice that different individual users have little chance to have a similar theme of tweeting about a specific topic. Collection of tons of Tweets in nearer geographical regions during specified time span may give measurable weight to the specific topic. This leads to the area of research. What is trending in Society by analyzing data of Microblogs? Starting from zilch, the attempt was more focused on finding trending terms. To identify trending terms from the mountain of Tweets, many hybrid approaches include information retrieval and machine learning techniques were adopted. The distillation process resulted in trending terms showing the reflection of real events of society. It is important to find out what is trending at this moment and not what has been in the discussion for a long time. From the perspective of social media analyst, one needs to define a long time. The term long time not only includes last month or last week but also encompasses last day or last few hours before the occurrence of the phenomenon in order to identify the trend of the latest moment. The approach used in analysis requires a lot of time and resources to process huge quantum of data. So, the basic concept is to breaking down the huge dataset and digging deep into the latest discussions only to produce trending terms of the moment, which is further tuned to by assigning enough weight to feedbacks from previous trending terms which ultimately produces more refined results. The concept of Trendiness-Distance has been proposed representing the importance of other terms with reference to trending terms. Trendiness-Distance of any term in different timespan is used to study the behavior of that term indicating the increasing or decreasing or steady graph with respect to the real trend. Merely single trending term will not lead us to an appropriate direction toward trend, but the set of trending terms collectively produce meaning toward trend. The proposed research work has been carried out to devise unique pairs from combinations of trending terms & identify maximum discussed trending pairs and thus adding more refined meaning to output. Concerning the much practical application of Trend, one can derive crucial information regarding specific product or service. One can devise new and innovative strategies for future challenges in order to serve more effectively to set of mass. Considering the tremendous importance of prediction usefulness, the work was mainly focused on predicting trendiness weight of given trending term for future cycles using the Poisson Distribution algorithm.