We just released LingPipe 3.8.1, which patches LingPipe 3.8.0. As usual, it’s available from the home page:
There are only two changes:
- Bug fixes/performance for logistic regression
We removed the object allocations introduced in 3.8 to reduce garbage collection overhead.
We fixed a bug introduced in 3.8 that prevented all but the first category in a multinomial logistic regression from being regularized.
- Start/End points for tokenizers
We added an end point method to tokenizers, and implemented start/end with unit tests for all of the supplied tokenizers. This makes for easier text highlighting, especially for filtered tokenizers like Porter stemming or stoplisting.