The release of v1.1.0 addresses a number of internal issues, including more comprehensive unittesting and fixing issues of function name override. A few new features have been added with support for custom keras
models and operators for Diagnostics
class. See the changelog and deprecation notice below for details.
There is no breaking change, and deprecations have minimal effects on end-users. This upgrade is recommended for all.
Changelog
- Added comparison operator support to the
Diagnostics
class - Added support for concatenating two
Diagnostics
instances using+
and+=
- Implemented the
Info
class as a singleton - Added support for using custom model and dictionary in the
Predictor
class - Added support for loading custom dictionary and model using the
Initializer
class - Changed "sent model" to "lexical model" for naming accuracy
- Added theoretical support for python 3.5 (No CI testing)
- Deprecated function parameters ("input" and "dict") to avoid overwriting builtin functions and methods
- Optimized unittest infrastructure with more test coverage
- Added Github README to
pypi
Deprecations
Some method parameters have been renamed:
- Renamed "input" to "lexical_input" in the following methods:
poetic.predictor.Predictor.predict
poetic.predictor.Predictor.preprocess
poetic.predictor.Predictor.tokenize
poetic.predictor.Predictor.word_id
- Renamed "dict" to "dictionary" in the
poetic.predictor.Predictor
constructor
Positional arguments remain unchanged. Deprecated arguments, "input" or "dict", become
keyword-only arguments for backwards compatibility if they have been explicitly named in
function calls.