This article describes how a rumor can be defined as a circulating unverified story or a doubtful truth. Rumor initiators seek social networks vulnerable to illimitable spread, therefore, online social media becomes their stage. Hence, this misinformation imposes colossal damage to individuals, organizations, and the government, etc. Existing work, analyzing temporal and linguistic characteristics of rumors seems to give ample time for rumor propagation. Meanwhile, with the huge outburst of data on social media, studying these characteristics for each tweet becomes spatially complex. Therefore, in this article, a two-fold supervised machine-learning framework is proposed that detects rumors by filtering and then analyzing their linguistic properties. This method attempts to automate filtering by training multiple classification algorithms with accuracy higher than 81.079%. Finally, using textual characteristics on the filtered data, rumors are detected. The effectiveness of the proposed framework is shown through extensive experiments on over 10,000 tweets.