Awesome

KCAT

A Knowledge-Constraint Typing Annotation Tool

Entity Linking

Download code and data. The model is based on Mulrel, we replace the source file "ed_ranker.py" with our file "ed_ranker.py" in directory "EL".

Train the model by

export PYTHONPATH=$PYTHONPATH:../
python -u -m nel.main --mode train --n_rels 3 --mulrel_type ment-norm --model_path model

Run the api by

python elapi.py --model_path model

Test the api in web

The url likes this, http://{address}:{port}/edl/ranking/?text={text}&&offset={offsets}, where address and port can be changed, text is the raw text and offsets are the spans of entity mention generate by Entity Discovery whose format is "b1-e1:b2-e2".

An example with top score entity is following.

Input:

http://10.214.155.248:5000/edl/ranking/?text=obama%20is%20america%20president&&offset=0-4

Result:

[{"start": 0, "end": 4, "surface_form": "obama", "entity": "Barack_Obama"}]

Fined-grain Typing

Requirements: Python 3.5 or 3.6, tkinter

run following command to start annotator client

python annotation.py --dataset dataset_name --url url_port