ACQuA
Answering Comparative Questions with Arguments
The goal of the ACQuA project (funded within the DFG-SPP 1999 RATIO: Robust Argumentation Machines ) is to develop algorithms and technology that help to understand and answer comparative information needs expressed as natural language questions by retrieving and combining facts, opinions, and arguments from knowledge graphs and web-scale text resources.
The project is motivated by the fact that everyone faces a variety of choices on a daily basis (e.g., what programming language to use or whether to buy an electric car) and often can easily formulate a respective question containing the potential options and important aspects.
Exploiting the web as a knowledge source, an answer to a comparative question should ideally directly combine the available facts, opinions, and arguments in a (short) natural language answer explaining under what circumstances which alternative should be chosen and why. This is the envisioned behavior of our comparative argumentation machine (CAM) for which we work on the following modules in the ACQuA project:
(1) a user-friendly interface to submit a comparative question in natural language,
(2) a question understanding component that identifies the compared objects and important comparison aspects,
(3) a system that retrieves appropriate facts from a knowledge graph and relevant (possibly argumentative) documents from a web-scale text resource,
(4) a component that generates a (short) natural language answer from the different extracted facts and retrieved documents.
ACQuA is a collaborative project with the "Language Techology" group from the Universität Hamburg.
[Demo ] Comparative Argumentation Machine (CAM).
[Demo ] TARGER: Neural Argument Mining.
We are organizing the second argument retrieval shared task Touché @ CLEF: Argument Retrieval .
[Publications ].