antannotator 0.0.3

JupyterLab tool for data annotations for machine learning routines

AntAnnotator

JupyterLab tool for data annotations for machine learning routines.

Just drop your python data into widged and label it.

Features

Data persistence Data is automatically persisted on disk in .json format.

UI Annotation UI witch allow to do annotation

Usage example

from antannotator.annotation_controller import AnnotationController
from antannotator.persistence import AutoEventStorage, AutoAnnotationTaskStorage, AnnotationSample

# Select directory where annnotation task data will be stored

event_storage = AutoEventStorage.get_event_storage("/tmp/annotator/events")
task_storage = AutoAnnotationTaskStorage.get_task_storage("/tmp/annotator/tasks")
controller = AnnotationController(task_storage=task_storage, event_storage=event_storage)

# 
options = {"literature": "Classical literature",
            "songs": "20th Century Songs", 
            "other":"Other"}


# load samples to annotate
samples = [ AnnotationSample(sample_id=f"id_1", 
            task_data=f"We passed upon the stair\nWe spoke of was and when\nAlthough I wasn't there\nHe said I was his friend", 
            available_options=options, 
            hidden_info=None,
            visible_info=None),
        AnnotationSample(sample_id=f"id_2", 
            task_data=f"It was many and many a year ago,\nIn a kingdom by the sea", 
            available_options=options, 
            hidden_info=None,
            visible_info=None),
        AnnotationSample(sample_id=f"id_3", 
            task_data=f"On a dark desert highway, cool wind in my hair\nWarm smell of colitas, rising up through the air", 
            available_options=options, 
            hidden_info=None,
            visible_info=None)]

controller.append_samples(samples)

# start annotations process
controller.do_annotaions()

Installation

pip install antannotator

Roadmap

• Multichoice text classification
  Multiclass, multichoise text classification

• One choice text classification
  One choice classification. (Task adavance immediately user celect a option)

• Free input
  Free user input. (For example for summarization task)