Keyword Update
Our list of keywords has gone through a substantive review. The most immediate news is that these changes will cause the model to behave in very strange ways for the next seven days. Next Friday about this time should provide users with a map with much better overall results.
Currently, we search for and aggregate 9,145 keywords. All of these keywords are designed to identeify country-level crises and discrete events. The original list of keywords was oriented towards identifying crisis events, but in different systems (ie., crisis events that could be traced to political systems, economic systems, etc.). That approach has been put on the shelf for the moment, as we do not have enough computing power to search for as many keywords as we would like. We are limited by our basic web hosting service to a certain amount of computer processing minutes a day, and we butt up against that nicely.
Because our country-level intensity is a reflection of an aggregation of keywords and a comparison of that hour’s news cycle churn with a measure of the previous 167 news cycles, our results will be funny until we have enough data to create a solid 7 day base case.
Eventually, we will build upon this discrete event model with a model that does incorporate different national systems; being able to highlight tension within distinct systems is a way of more clearly identifying potential tipping points. These additional systmes will not be entirely comprised of media churn, but other kinds of churn as well (poll numbers, stock movement, currency valuations, etc).