Data Annotator and Image Labeler
The project involved annotating and labeling over 500,000 tweets to support Natural Language Processing (NLP), sentiment analysis, and machine learning model training. Tweets varied widely in style, ranging from formal and conversational to slang-heavy content with emojis, hashtags, abbreviations, and code-switching across languages like English, Hindi, and Japanese. Key tasks included sentiment classification, intent and topic labeling, entity recognition, and content moderation, along with metadata tagging for elements such as emojis and URLs. This ensured the creation of context-aware, comprehensive datasets for advanced NLP applications. Quality assurance was maintained through a multi-layered system. Annotators received detailed guidelines, training, and qualification tests, with each tweet reviewed by two to three independent workers. These measures consistently upheld accuracy benchmarks above 90%. Disagreements were resolved by majority vote or expert reviewers.