A simple script that shows how Orolo compares to chrono and recognizers-date-time and hopefully explains why orolo was needed.
$ npm install
$ node index.jsAll tests are in benchmark.json.
To get as close to "apples to apples" comparison as possible, I wrapped all engines under benchmark with converters that process "native" outputs into a collection of dates like orolo does.
Orolo: 76%
Chrono: 24%
Microsoft: 12%
To get a detailed report and see what the engines guess wrong you will have to run the script.
I am sure one can come up with the benchmark that yields quite the opposite results. This benchmark is representative of what kinds of dates and date ranges I needed to be able to recognise in a natural language sentence and that's why I built orolo in the first place.