This is the first release of Salience 6.1.
Features
- Queries can now be used in an hsd file to assign sentiment/
- Entity Sentiment is now calculated with a chunk parsing based algorithm. The old algorithm can still be used with the Chained Entity Sentiment option. The new algorithm is more accurate, but produces more neutral results.
- Queries are now optionally insensitive to accents/diacritics.
Improvements
- Pattern Files now support a (=~s~'term') syntax to match all stems of a term.
- Added suppport for multiple classification models at once.
- Improved negation handling
- Improved stemming of irregular English verbs.
- Language detection now supports Russian, Arabic and Dutch
- Added Use Chainer option which improves runtime performance at a small accuracy cost.
- Added support for Intentions with no objects, and some additional types in the pattern restriction file.
- Sped up Demo startup time
- Autocategories updated to match the current category naming on Wikipedia
- Some general performance improvements
Bug fixes
- Fixed support of multibyte characters in queries
- Fixed some occasional crashes
- Fixed issue with changing query options (like max query length) after having already loaded the query file.
- Fixed abbreviations in language detection: ge => de, sp => es
- If you have any questions, please email [email protected]
Updated 11 months ago