This tool allows users to labelize data one by one with a tkinter UI. You just need to give a set of label types and a set of data to be labelized. The interface will adapt itself. All labels will be stored either in a pickle file or in a mongo collection using a SerializableDict.

This tool is actually in beta, it works but a lot of functionnalities has to be implemented on the fly according to your needs.

git clone https://github.com/hayj/Annotator.git
pip install ./Annotator/wm-dist/*.tar.gz

First import the lib:

from annotator.annot import *

Get your data:

data = \
[
	{"author": "John", "text": "This is the first document." + " X" * 300},
	{"author": "Henry", "text": "The second doc." + " Y" * 30},
	{"author": "Roger", "text": "The third doc." + " Y" * 30},
]

Define a function which will yield all documents as dicts with id (must be unique) and content which will contain the data you want to display. The content field can be a list to have mulitple columns in the UI.

def dataGenerator():
	for current in data:
		author = current["author"]
		text = current["text"]
		yield {"id": author, "content": {"title": author, "text": text}}

Define your labels:

labels = \
{
	"relevance": {"title": "Relevance", "type": LABEL_TYPE.scale, "from": 0, "to": 100, "resolution": 10, "default": 50},
	"weight": {"title": "Weight", "type": LABEL_TYPE.scale, "default": 0.2},
	"format": {"title": "The document is well formed", "type": LABEL_TYPE.checkbutton, "default": True},
	"popular": {"title": "The author is popular", "shorttitle": "Popular author", "type": LABEL_TYPE.checkbutton},
}

Init the Annotator. The first argument is the name of the file or the name of the mongo collection. Here we store all annotations in a pickle file but you can also use mongodb if you set host, user, and password:

an = Annotator("my-annotations", labels, useMongodb=False, dirPath="/home/hayj/tmp")

Here you start the UI, so you will have one data on the left and labels on the right to be manually edited. You can click the right arrow button to switch to the next data, or return to the previous.

an.start(dataGenerator())

Screenshot of the UI:

Finally print your labeled data:

print(list(an.data.items()))