We are pleased to announce the IGGSA Shared Task on German Sentiment Analysis. It is open to any interested participants from both academia and industry.
Introduction Sentiment analysis and opinion mining have caught increasing attention both in the scientific community as well as in industry. Use cases include the analysis of products, marketing campaigns, brands, companies, political speeches and more. However, most work has been performed on the English language. With this shared task, we put a focus on the German language, both for analysing product reviews as well as political speeches in a fine-grained manner.
Task Description The shared task consists of two main tasks:
Main Task 1: Sentiment Analysis on German speeches from the Swiss parliament
Full task: Identification of subjective expressions with source and targets.
Subtask 1a: Subjective expressions are given, the task is to identify opinion sources.
Subtask 1b: Subjective expressions are given, the task is to identify the opinion targets.
Main Task 2: Cross-Domain Sentiment Analysis on German product reviews from Amazon.
Subtask 2a: Identify subjective phrases.
Subtask 2b: Identify aspect phrases.
Subtask 2c: Identify subjective phrases and aspect phrases and indicate for each aspect phrase which subjective phrase it is the target of (if any).
More detailed information can be found at https://sites.google.com/site/iggsasharedtask/
Data sets All tasks are evaluated on unseen test data. For main task 1, the data set consists of 600 sentences representing continuous segments of 25 speeches on 9 different topics given before the Swiss parliament. For main task 2, the test data will be in the same format as the USAGE data set, which serves as training data (it consists of annotations for more than 600 reviews, http://dx.doi.org/10.4119/unibi/citec.2014.14). The test data will be from a different product category than the ones included in the USAGE data set.
Schedule Main Task 1: May 1st, 2014: Release of Trial data August 1st, 2014: Release of evaluation data for main subtask August 10th, 2014: Submission of results (max. 5 runs) for subtask a August 11th, 2014: Release of data for subtask a and b August 21st, 2014: Submission of results (max. 5 runs) for subtask a and b September 1st, 2014: Report results to each team individually
Main Task 2: Now: Release of full training data August 11st, 2014: Release of evaluation data for all subtasks. August 21th, 2014: Submission of results (max. 5 runs) September 1st, 2014: Report results to each team individually
Both Main Tasks: July 15th, 2014: Registration deadline September 22nd, 2014: Working note submission Camera ready papers will be submitted after workshop, on October 30th, 2014.
Registration July 15th, 2014: Registration deadline Please register at: https://sites.google.com/site/iggsasharedtask/registration Every participant is strongly encouraged to register on our mailing list at https://groups.google.com/d/forum/iggsa-sentiment-analysis-shared-task
Workshop The workshop for the shared task will take place in Hildesheim/Germany on October 7th, 2014, one day prior to the beginning of the KONVENS conference, with which the workshop is co-located. (http://www.uni-hildesheim.de/konvens2014/).
Participants are asked to register for the main conference; the registration fee covers the main conference and all workshops.
Coordinators Josef Ruppenhofer (Uni Hildesheim) Jonathan Sonntag (Uni Potsdam) Julia Maria Struss (Uni Hildesheim) Michael Wiegand (Uni Saarland) Roman Klinger (Uni Bielefeld)
Contact and Mailinglist You can contact the organizers via the email address iggsa-gestalt-coordinators at googlegroups.com. In addition, we maintain a mailing list at https://groups.google.com/d/forum/iggsa-sentiment-analysis-shared-task for exchange between participants and organizers. We strongly encourage you to register on this list when you are interested in participation.
Acknowledgements We are happy to acknowledge the financial support that the GSCL (German Society of Computational Linguistics) has granted us for annotating task 1 data. Task 2 has been supported by the BMBF project It’s OWL (http://www.its-owl.de/).