Sentiment140 Tweet Sentiment Dataset Analyzed
A Twitter dataset of 1,600,000 tweets with binary sentiment labels (positive/negative) assigned via emoticon distant supervision, collected by Go, Bhayani and Huang at Stanford University via Twitter streaming API in 2009. Each record contains tweet text with polarity label (0=negative, 4=positive), tweet ID, timestamp, search query used for collection, and username. The dataset is notable for its one of the largest and most widely used Twitter sentiment benchmarks, notable for using emoticons as a noisy but scalable distant supervision signal rather than manual annotation, enabling 1.6M labeled examples. Intended for research into sentiment analysis, opinion mining, distant supervision research, text classification benchmarking, and social media language modeling.
Twitter (X)