Clément Doumouro
Clément Doumouro
# Current behavior Currently the `LogRegIntentClassifier` uses all the CPU on the machine it's training while the `ProbabilisticIntentClassifier` trains it's `CRFSlotFiller`s sequentially. # Expected behavior Ideally: - the `CRFSlotFiller` training...
# Problem description ## Short description In certain conditions some CRF tags transitions can by missing after the data augmentation or can be "underrepresented". We must ensure that all possible...
# Parsing Error ## Version `your.tag.number` ## Language `EN` ## Parser input `one two three` ## Parser output ```python parser = RustlingParser(u"en") parser.parse(u"one two three") >>> [{'char_range': {'end': 13, 'start':...
Depends on #1079, #1081 and #1283 At the end of this story, BatchSearchLoop and BatchDownloadLoop should be the same. * [x] do #1283 to make it easier for this story...
Currently Tika forgets about soft line breaks detected inside PDF, these line break shouldn't be translated into hard line breaks inside Datashare.
Ideally, DS CLI should offer a way to run commands and access their help in a hierarchical manner, similar to the AWS CLI or other large CLI apps: ```bash datashare...
https://github.com/CodeStory/fluent-http/issues/162
# Description - [x] create a evaluation dataset - [x] create an evaluation pipeline - [ ] train a first version of the pipeline without dedupe in a nomenklatura style...
**Is your feature request related to a problem? Please describe.** Currently as a developper it's not possible to create tasks from a Datashare extension: - the `TaskManager` interface is not...