Image Retrieval/Generation for Arguments 2024

Argumentation is a communicative activity of producing and exchanging reasons to support claims. Most of the time, this is done abstractly with words. This has the advantage that we can formulate exactly, but we need time and effort to understand the words. Images, on the other hand, affect us directly. Look at the example below; what gets your attention first? There is a high chance that you looked at the images before reading the involved argument. Also, in the pictures, you can see directly how many nuns were involved and what the voting station looks like, though this is not stated in the written argument. In this regard, images are less precise than words but contain more information. For this task, we want to use the power of images for argumentation.

Synopsis

  • Task: Given an argument, find images that help to convey the argument’s premise.

“Convey” is here meant in a general manner; it can depict what is described in the argument, but also show a generalization (e.g., a meme image that illustrates a related abstract concept) or specialization (e.g., a concrete example).

For this task, an argument consists of one claim and a premise. In addition, we provide the argument's topic and the premise's type. Premises are either facts from a study or anecdotal evidence and are labeled accordingly so that participants can use different approaches for these types.

Submission

We allow three kinds of submissions. Each of the submission styles is displayed in the example.

  1. Retrieval. Like in the last years, participants can retrieve suitable images from a focused crawl, where we also provide automatically recognized text from the image (OCR) and text from web pages that contain the image.
  2. Prompted Generation. Following the idea of the infinite index, participants can submit prompts for the Stable Diffusion image generator.
  3. Direct. Participants can employ other reproducible methods for generating images and directly submit them. This includes chart generators, which can generate a bar chart from given numbers in the premise. Also, one can use headline generators to transform the premise into a headline.

Data

We provide access to a Stable-Diffusion API for image generation. For participants favoring image retrieval, we provide access a focused crawl of about 10K images (and associated web pages) as document collection.

Example

Premise: Indiana’s photo ID law barred twelve retired nuns in South Bend, Indiana from voting in that state’s 2008 Democratic primary election. The women lacked the photo IDs required under a state law that was upheld by the U.S. Supreme Court in April 2008

Claim: Legislation to impose restrictive photo ID requirements has the potential to block millions of eligible American voters, and thus suppress the right to vote

Topic: This house believes that democratic governments should require voters to present photo identification at the polling station>

Type: Anecdotal

Retrieved/Generated Images:

self generated image.

Self-generated headline image

generated by Stable Diffusion.

Image generated by Stable Diffusion

web retrieval.

Image retrieved from the web

Task Committee