v2.0.1

• Switched away from pyramid ARIMA due to stability issues
• Now supports raw and bzip'd pickle files
• Using Python 3.7.3
• Handles imported data when dates are stored as strings rather than Timestamp objects
• Corrected handling of unigrams

Emergent Terms

• New: Detects emergent terms as well as popular terms
• New: Term weightings using word vectors
• New: Prediction of future term counts

FDG bug fix

Force directed graph URL for empty.json was not updated when folder structure was reorganised

JSON config support

Added support to create JSON config file, recording configuration used to produce report.

Table output and enhanced CI

• New: table format output
• New: V & V code added
• New: CI with MacOS (via Travis)
• New: CI with Windows (via AppVeyor)
• Various documentation fixes
• Missing attributions corrected

Updated and corrected documentation

Updated README.md and tidied FDG code to simplify HTML / JS output

1.0.0 release

First release of app_detect; this demonstrates analysis of patent abstracts to determine popular terminology, which can be presented as a text report, word cloud or force-directed graph. This has been tested under MacOS 10.13.6.

Draft release to test process

Draft release; integrated into Travis and Codecov, all tests passing.
detect.py analyses pickled data frames of patent data, supplied with data generated from USPTO publicly available datasets.