Please use this identifier to cite or link to this item:
|Title:||Semantic Computation of Twitter Hashtags|
|Abstract:||Hashtags is a metadata or label followed by a “#” sign, it may be a word or a phrase that is used on SNS especially frequently on Twitter. People use hashtags in tweets for getting popularity, trending, advertising, promotion, searching specific topic etc. is really hashtags represent the Tweets? For this purpose, we develop a system that finds deeply semantic analysis & relatedness / similarity b/w hashtags and tweets. Our Web system works in three parts: Extraction of Tweets, which uses Twitter Rest API to extract Tweets from Twitter Web Site, second Separation and segmentation, which separate hashtags from tweets, uses different methods to segmentation of hashtags and also separate part of speech (POS) from tweets and third semantic relatedness calculation which finds the semantic similarity / relatedness among segmented hashtag words and POS by using different ten algorithms. After calculation of semantic relatedness, we find the arithmetic mean of that score. Adapted Extended Lesk’s measurements strongly shows that hashtags represents the tweets with mean 0.701, also some other algorithms’ measurements admits the statement with mean Leacock & Chodorow 0.456, Resnik (Res) 0.392, Hirst & St-Onge (HSO) 0.338, Jiang & Conrath (Jcn) 0.324 and Wu & Palmer (wup) with score 0.232. We also done User Evaluation, four users gives their feedback either hashtags represents the tweets or not in the form of yes / No by evaluating all tweets data set. After feedback we use Fleiss' kappa calculation to find the degree of agreement among users’ feedback, the calculation of Fleiss' kappa is 78.22% that also shows that tweets are represented by hashtags, and have major role in popularity of tweets.|
|Appears in Collections:||Thesis|
Files in This Item:
|Semantic Computation of Twitter Hashtags.pdf||762.74 kB||Adobe PDF||View/Open|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.