Sprachen

New Publications

Towards Linked Research Data: An Institutional Approach
Bi-directional Inter-dependencies of Subjective Expressions and Targets and their Value for a Joint Model
Committee-based Selection of Weakly Labeled Instances for Learning Relation Extraction
Proceedings of Interacting with Linked Data (ILD 2012), workshop co-located with the 9th Extended Semantic Web Conference, May 28, 2012, Heraklion, Greece

More...

Benutzeranmeldung

QALD-1

1st Workshop on Question Answering over Linked Data (QALD-1)

May 30, 2011, Heraklion, Greece (co-located with the ESWC 2011)

Organizers:

Christina Unger (CITEC, Bielefeld University)
Philipp Cimiano (CITEC, Bielefeld University)
Vanessa Lopez (KMi, Open University)
Enrico Motta (KMi, Open University)

Endorsed by:

Invited speaker:

Chris Welty (IBM Research)

Description:

While more and more semantic data is published on the Web, the question of how typical Web users can access this body of knowledge becomes of crucial importance. So far there are not many paradigms which would allow end users to profit from the expressive power of these standards while at the same time hiding the complexity behind an intuitive and easy-to-use interface. Important and promising research directed towards this goal is offered by search paradigms based on natural language question answering, which allow users to express arbitrarily complex information needs in an intuitive fashion. The essential problem lies in translating these information needs into a form such that they can be evaluated using standard Semantic Web query processing and inferencing techniques.

In recent years, there have been important advances in semantic search and question answering over RDF data. In parallel to these developments in the Semantic Web community, there has been substantial progress on question answering from textual data as well as in the area of natural language interfaces to databases. An important challenge for the Semantic Web, but also Natural Language Processing communities, is scaling question answering approaches to Linked Data. The main challenges involved herein are related to dealing with a heterogeneous, distributed and huge set of interlinked data.

Our goal is to bring together research and expertise in question answering from different communities, including NLP, HCI, Semantic Web and Databases. To facilitate the comparison between different approaches and systems, and to foster research in this area, the workshop is accompanied by an open challenge (see below).

The following topics are of special interest:

Question analysis
Natural language processing approaches applied to semantic search engines
Disambiguation and inferencing across multiple sources and domains
Distributed evaluation of queries over Linked Data
Lexical resources supporting question answering over Linked Data
Discovery on the fly of relevant Linked Data sources
Support for multiple languages
Methods for scaling up question answering to Linked Data
Efficiency and performance aspects
Dealing with data and schema heterogeneity
Answer merging and ranking
Providing justifications of answers and conveying trust
User feedback and interaction
Habitability and usability aspects
Evaluation of question answering approaches for Linked Data

Workshop Program (May 30, 2011)

09:00 - 10:00	Chris Welty (Invited speaker):
	Inside the mind of Watson
10:00 - 10:30	Damljanovic, Agatonovic & Cunningham:
	FREyA: an Interactive Way of Querying Linked Data using Natural Language

10:30 - 11:00	Coffee break

11:00 - 11:30	Freitas, Oliveira, Curry, O'Riain & Pereira da Silva:
	Treo: Combining Entity-Search, Spreading Activation and Semantic Relatedness for Querying Linked Data
11:30 - 12:00	Granberg & Minock:
	A Natural Language Interface over the MusicBrainz Database

12:00 - 12:30	Demo session
12:30 - 13:00	Discussion

Proceedings

NEW: proceedings.pdf

Open Challenge

To facilitate the comparison between different approaches and systems, the workshop was preceded by an open challenge in question answering on DBpedia 3.6 and MusicBrainz.

Full description: sharedtask.pdf (last updated: March 28)

The DBpedia dataset can be downloaded here:

An RDF export of relevant MusicBrainz data can be downloaded here:

musicbrainz.tgz (126.4 MB)
NOTE: There was an error in musicbrainz.owl. The repaired file is in the archive and here. (Last updated: February 20)

You can also access the data via a SPARQL endpoint.

For each dataset, there are 50 training questions of different complexity levels with manually specified SPARQL queries and answers:

dbpedia-train.xml (HTML encoding)
dbpedia-train-CDATA.xml (CDATA encoding, not up-to-date)
musicbrainz-train.xml (HTML encoding)
musicbrainz-train-CDATA.xml (CDATA encoding)

Participating systems were evaluated with respect to precision and recall on 50 similar test questions:

All question are provided in English.

For both the training and the test phase, there was an online evaluation form, where results could be uploaded: http://greententacle.techfak.uni-bielefeld.de/~cunger/qald1/evaluation
Upload of results for test questions was possible from April 1 to April 10.

For updates, please subscribe to the QALD-1 mailing list.

NEW: Evaluation results (test phase)

NEW: Gold standard (test phase):

Call for Papers (PDF)

We welcome the submission of research papers on all aspects of question answering over Linked Data, system descriptions of systems participating in the shared task, and short papers on preliminary research.

Schedule:

Release of training dataset	January 31
Release of test dataset	March 28
Submission of results	April 1 - April 15
~~Deadline for workshop papers~~	~~March 4~~
Deadline for abstracts (REQUIRED)	March 4
Deadline for full papers (EXTENDED)	March 14
Notification of acceptance	March 30
Camera-ready version of papers	April 22
Workshop	May 30

Submission Details

Research papers should not exceed 12 pages in length (excluding references), system descriptions should not exceed 8 pages (excluding references). Short papers describing preliminary research or positions can also be submitted and should not exceed 4 pages in length. All papers have to be submitted in PDF format; camera-ready versions must be formatted according to the LNCS guidelines.
Submission and reviewing will be via EasyChair:

http://www.easychair.org/conferences/?conf=qald1

Program committee: