Introduction In an era where information is relentlessly penetrating people’s minds and shaping their opinion and their actions leading in some occasions to strife and societal dysfunction, tools to provide adequate awareness for readers has become crucial for the well being of a society. The challenge of determining the truth is an age-old battle humans have relentlessly pursued whose solution could require immense resources (source verification, source trust, observation objectivity...) and so needs further investigation. We, on the other hand, have decided to build a tool, using Google’s Natural Language Processing sentiment analysis on its cloud platform to judge an online article, given its URL (or custom text) based on its writing. It would allow users to know, before reading an article, about how much it is emotionally charged or potentially trying to influence them (whether to the positive or negative). Distinguish between strong opinions and logical reasoning. This tool would ideally decrease the severity of the psychological effect of extreme articles and reinforcement bias, while potentially adding value to articles that rely more on reason. Technical Challenges: First time using Google cloud and their natural language processing (NLP) API Determining threshold to classify documents (manual)
- Select representative sample of online articles 2. Run NLP and extract parameters 3. Plot parameter histogram 4. Decide on classification thresholds Limitations: sometimes incorrect “facts” without explicit emotional/exaggerating statements could still influence individuals. Next Step: determine negative/positive direction of emotionally charged articles.