Skip to main content

Library for semantic annotation of tabular data with the Wikidata knowledge graph

Project description

bbw (boosted by wiki)

PyPI version badge badge Language grade: Python

  • Annotates tabular data with the entities, types and properties in Wikidata.
  • Easy to use: bbw.annotate().
  • Resolves even tricky spelling mistakes via meta-lookup through SearX.
  • Matches to the up-to-date values in Wikidata without the dump files.
  • Ranked in third place at SemTab2020.

Table of contents

How to use

Import library

from bbw import bbw

The easiest way to annotate the dataframe Y is:

[web_table, url_table, label_table, cpa, cea, cta] = bbw.annotate(Y)

It returns a list of six dataframes. The first three dataframes contain the annotations in the form of HTML-links, URLs and labels of the entities in Wikidata correspondingly. The dataframes have two more rows than Y. These two rows contain the annotations for types and properties. The last three dataframes contain the annotations in the format required by SemTab2020 challenge.

The fastest way to annotate the dataframe Y is:

[cpa_list, cea_list, nomatch] = bbw.contextual_matching(bbw.preprocessing(Y))
[cpa, cea, cta] = bbw.postprocessing(cpa_list, cea_list)

The dataframes cpa, cea and cta contain the annotations in SemTab2020-format. The list nomatch contains the labels which are not matched. The unprocessed and possibly non-unique annotations are in the lists cpa_list and cea_list.

GUI

If you need to annotate only one table, use the simple GUI:

streamlit run bbw_gui.py

Open the browser at http://localhost:8501 and choose a CSV-file. The annotation process starts automatically. It outputs the six tables of the annotate function.

Try it out online (no SearX support) with this binder link.

CLI

If you need to annotate a few tables, use the CLI-tool:

python3 bbw_cli.py --amount 100 --offset 0

GNU parallel

If you need to annotate hundreds or thousands of tables, use the script with GNU parallel:

./bbw_parallel.py

Installation

You can use pip to install bbw:

pip install bbw

The latest version can be installed directly from github:

pip install git+https://github.com/UB-Mannheim/bbw

You can test bbw in a virtual environment:

pip install virtualenv
virtualenv testing_bbw
source testing_bbw/bin/activate
python
from bbw import bbw
[web_table, url_table, label_table, cpa, cea, cta] = bbw.annotate(bbw.pd.DataFrame([['0','1'],['Mannheim','Rhine']]))
print(web_table)
deactivate

Install also SearX, because bbw meta-lookups through it.

export PORT=80
docker pull searx/searx
docker run --rm -d -v ${PWD}/searx:/etc/searx -p $PORT:8080 -e BASE_URL=http://localhost:$PORT/ searx/searx

SearX is running on http://localhost:80. bbw sends GET requests to it.

Citing

If you find bbw useful in your work, a proper reference would be:

@inproceedings{2020_bbw,
  author    = {Renat Shigapov and Philipp Zumstein and Jan Kamlah and Lars Oberl{\"a}nder and J{\"o}rg Mechnich and Irene Schumm},
  title     = {bbw: {M}atching {CSV} to {W}ikidata via {M}eta-lookup},
  booktitle = {SemTab@ISWC 2020},
  year = {2020}
}

[preprint] [presentation] [BERD@BW]

SemTab2020

The library was designed, implemented and tested during SemTab2020. It received the best scores in the last 4th round at automatically generated dataset:

Task F1-score Precision Rank
CPA 0.995 0.996 2
CTA 0.980 0.980 2
CEA 0.978 0.984 4

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for bbw, version 0.1.2
Filename, size File type Python version Upload date Hashes
Filename, size bbw-0.1.2-py3-none-any.whl (19.7 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size bbw-0.1.2.tar.gz (27.8 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page