Johannes Hoffart

I am a research scientist at Goldman Sachs in the area of Natural Language Understanding and Knowledge Graphs. In the past, I co-founded two startups (Ambiverse and nudge:nudge) with the goal of creating machines that organize knowledge with AI methods. I did my PhD at the Max Planck Institute for Informatics, focusing on discovering and disambiguating named entities in text.

Have a look at my publications and projects.

Follow @johanneshoffart

What I Do

Research

My research interest is in natural language understanding, knowledge base construction, and web-scale knowledge harvesting.

Invited Talks & Tutorials

Natural Language Understanding and Knowledge Bases in Finance (and AI Research at Goldman Sachs).
NEC Labs Heidelberg, Heidelberg, Germany, January 2019
Riesige Mengen von Dokumenten besser durchsuchen und analysieren.
Bundeszentralamt für Steuern, Bonn, Germany, February 2017
Knowledge Bases for Web Content Analytics.
Tutorial at the 24th International World Wide Web Conference, Florence, Italy, WWW 2015
Discovering and Disambiguating Named Entities in Text.
Fudan University, Shanghai, China. November 2014
Discovering and Disambiguating Named Entities in Text.
HPC Lab at ISTI, CNR, Pisa, Italy. October 2014
Constructing a Spatio-Temporal Knowledge Base and using it to Process and Understand Natural Language Text.
Seoul National University (invited by Saltlux), Seoul, South Korea. August 2013
Robust Disambiguation of Named Entities in Text.
Google Zurich, Switzerland. August 2011

Publications

[bibtex file=mypubs.bib groupByYear=true]

Projects

AIDA & AmbiverseNLU

AIDA is a framework and online tool for entity detection and disambiguation. Given a natural-language text, for example news articles, it maps mentions of ambiguous names onto canonical entities (e.g., individual people or places) registered in the YAGO knowledge base. A description and demo is available on the AIDA website. AIDA is now part of the AmbiverseNLU suite.

AIDA Example — Finding the meaning in a sentence with AIDA.

YAGO

YAGO is a huge semantic knowledge base, derived from Wikipedia, WordNet and GeoNames. Currently, YAGO has knowledge of more than 10 million entities (like persons, organizations, cities, etc.) and contains more than 120 million facts about these entities. All the data as well as several interfaces to browse and query the data are available on the YAGO website.

STICS

STICS is an entity-centric search engine that makes use of AIDA and YAGO. By extending the Google slogan of “things, not strings” to support also entity categories, STICS provides powerful functionality for querying and analyzing news and other text corpora in terms of entities, semantic classes, and text phrases. You can search, for example, for presidents of the United States and the JFK airport, and see how STICS distinguishes between JFK and JFK.