QR2020: Using Orders Of Magnitude Reasoning To Aggregate And Compare News Reporting Sentiment

This interesting paper examines sentiment analysis using GDELT.

This paper focuses on analyzing the underlying sentiment of news articles, taken to be factual rather than comprised of opinions. The sentiment of each article towards a specific theme can be expressed in orders of magnitude terms and aggregated into a centralized sentiment which can be trended. This allows the interpretation of sentiments without conversion to numerical values. The methodology, as defined, maintains the range of sentiment articulated in each news article per day. In addition, a measure of consensus is defined for each day as the degree to which the articles published agree in terms of the sentiment presented. A real case example is presented for a controversial event in recent history with the analysis of 82,054 articles over a three day period. The analysis compares the internal consensus per day of different countries.

Read The Full Article.