Finding Records in Social Media: A Natural Language Processing Fundamentals Exploration

Social media postings are now routinely used as proof of activities, events, or transactions in news media, academic institutions, governments, judicial courts, commerce, and various other organizations. The need to preserve social media content as records has drawn the interest of academic researchers, industry professionals, and policy makers. Despite the importance of this research area, selection of records from a pool of social media content remains an area of low research activity. This paper explores the use of Natural Language Processing methods to classify and select records from a pool of tweets (twitter social media content). We experiment with various characteristics of the data and NLP parameters with the goal of determining optimal parameters for training a supervised machine learning classifier. This paper can serve as an aid for understanding the fundamental elements of automating the selection of social media records