AIDA: Accurate Online Disambiguation of Named Entities in Text and Tables

Overview STICS Downloads

News

[April 2, 2014]
Released AIDA-light source code. AIDA-light will be presented at the Linked Data workshop at WWW 2014.
[April 2, 2014]
Released AIDA-EE dataset. The dataset is part of the publication J. Hoffart, Y. Altun, and G. Weikum: Discovering Emerging Entities with Ambiguous Names. WWW 2014.
[March 10, 2014]
Presenting "STICS: Searching with Strings, Things, and Cats" at CeBIT 2014 in Hannover.

Stay up to date with AIDA news and releases, send a mail to: aida-news-subscribe@lists.mpi-inf.mpg.de

Overview

AIDA is a framework and online tool for entity detection and disambiguation. Given a natural-language text or a Web table, it maps mentions of ambiguous names onto canonical entities (e.g., individual people or places) registered in the YAGO2 knowledge base.

You can try AIDA on any text you like in the online demo. In case you are interested in using AIDA programmatically, the source code is available on github.com/yago-naga/aida.

To experimentally verify the quality of AIDA, we annotated nearly 1,400 newswire articles with the entities mentioned in each article. This collection is available for download (see Downloads).

Download AIDA

The AIDA source code is available on github.com/yago-naga/aida. For AIDA to work, you will need to download our YAGO-based entity respository from our download page and import it into a PostgreSQL server.

Datasets

Find all datasets related to AIDA in our downloads area.

Publications

Discovering Emerging Entities with Ambiguous Names
Johannes Hoffart, Yasemin Altun, Gerhard Weikum
In: Proceedings of the 23rd International World Wide Web Conference, p. 385–395. WWW 2014, Seoul, South Korea, 2014
AIDA-light: High-Throughput Named-Entity Disambiguation
Dat Ba Nguyen, Johannes Hoffart, Martin THeobald, Gerhard Weikum
In: Linked Data on the Web, WWW 2014, Seoul, South Korea, 2014
KORE: Keyphrase Overlap Relatedness for Entity Disambiguation
Johannes Hoffart, Stephan Seufert, Dat Ba Nguyen, Martin Theobald, and Gerhard Weikum
In: Proceedings of the 21st ACM International Conference on Information and Knowledge Management, p. 545-554, CIKM 2012, Maui, USA, 2012
AIDA: An Online Tool for Accurate Disambiguation of Named Entities in Text and Tables PDF
Mohamed Amir Yosef, Johannes Hoffart, Ilaria Bordino, Marc Spaniol, Gerhard Weikum
In: Proceedings of the 37th International Conference on Very Large Databases, VLDB 2011, p. 1450–1453, Seattle, WA, 2011
Robust Disambiguation of Named Entities in Text PDF
Johannes Hoffart, Mohamed Amir Yosef, Ilaria Bordino, Hagen Fürstenau, Manfred Pinkal, Marc Spaniol, Bilyana Taneva, Stefan Thater, Gerhard Weikum
In: Conference on Empirical Methods in Natural Language Processing, p. 782–792, Edinburgh, Scotland, 2011
For scientific works, please cite this paper

Demo

AIDA can be tested online, with different methods and configurations for entity disambiguation: AIDA Demo

Please use either Firefox or Chrome to view the Demo.

Results

In the EMNLP 2011 paper, the results are given on the subset of 228 CoNLL testb documents wich could be processed by all the competitor methods (1270testb, 1308testb, and 1349testb are missing). We did this for the sake of comparability. The results of our AIDA methods on all 231 CoNLL testb documents are given below. The short names are the same as in the paper.

AIDA results in % on all 231 CoNLL-YAGO testb documents (using the original EMNLP 2011 code)
Measure	sim-k	r-prior sim-k	r-prior sim-k coh	r-prior sim-k r-coh	prior
Macro Precision@1.0	76.65	80.81	80.86	82.02	71.36
Micro Precision@1.0	76.65	80.06	82.24	82.29	66.55

AIDA results in % on all 231 CoNLL-YAGO testb documents (using the code re-engineered early 2013)
Measure	sim-k	r-prior sim-k	r-prior sim-k coh	r-prior sim-k r-coh	prior
Macro Precision@1.0	76.00	81.03	80.67	81.66	75.16
Micro Precision@1.0	76.61	80.56	82.05	82.54	70.46

Further Information

If you need any further information, please contact us via mail: NAME_OF_PROJECT@mpi-inf.mpg.de