Lettria

Added
- New feature: Pattern Matcher
- New feature: Dependency Pattern Matcher
- New function (lettria.load_results) that can return an iterator of NLP objects for reading files chunk by chunk.
Changed
- Default save format changed from json to jsonlines, this allows to load huge files in memory by reading them chunk by chunk
- Default representation for TextChunk object has changed from raw json data to token text.
- Important decrease of memory requirements and results file size (Around -70%).

Added
- Sentiment methods are now accessible through the functional interface (get_sentiment(), word_sentiment() etc.).
- Methods list_entities(), word_frequency(), word_count(), statistics(), and vocabulary() are accessible through the functional interface.
- split_results() to split a result file into multiple files
- filter_type() for filtering sentences based on their type (command, assert, question_open, question_closed).
- pos_detail, lemma_detail properties accessible at all levels through functional interface.
Changed
Deprecated
- Sentiment module is deprecated and will be removed in a future release.
Removed
- Methods list_entities(), word_frequency(), word_count() and vocabulary() had their 'level' argument removed.
The 'level' is now chosen by using the method at the approriate level
Before: self.nlp.vocabulary(level='documents') => After: [doc.vocabulary() for doc in self.nlp]
Fixed
- pos_detail property for Token class has been fixed.
Security

Added
- SDK redone from the ground up ! Look at our tutorials and documentation to get started.
- New interface with the class NLP that provides an easy access to relevant results.
- New document system that allows you to classify and name your data the way you want.
- Sentiment class provides additional functionalities for sentiment and emotion analysis
- Easy to save and load results from the API.
Changed
Deprecated
Removed
Fixed
Security

Added
- Analyser class to provide easy access and analysis of API result..
Changed
Deprecated
Removed
Fixed
Security