Kudos to my collaborators @srishtiy.bsky.social (who co-led this work), @mariaa.bsky.social @serge.belongie.com @iaugenstein.bsky.social for making this happen!

In "Investigating Human Values in Online Communities", we perform a high-scale study of the unique values expressed by online communities with different perspectives
arxiv.org/abs/2402.14177 #NAACL2025 #NLProc
@nadavb.bsky.social @rnv.bsky.social @frimelle.bsky.social @iaugenstein.bsky.social

C3NLP Workshop #NAACL2025:
@iaugenstein.bsky.social will be presenting on tailoring LLM outputs to cultures, including where implicit cultural personalisation based on names leads to over-simplification arxiv.org/abs/2502.11995
@rnv.bsky.social @frimelle.bsky.social
bsky.app/profile/frim... #NLProc

The high effort solution is to use an LLM to make a browser extension which tracks your academic reading and logs every paper you interact with to github, which builds and publishes a webapp to expose the data.

Which, clearly only a crazy weirdo would do.

dmarx.github.io/papers-feed/

Using simple, small models with the goal of usability and scalability of the task, we hope social scientists, journalists and researchers use this as a first step in studying multimodal framing and its intended/unintended effects.

More here:

bsky.app/profile/mari...

Find other interesting results in our paper: arxiv.org/abs/2503.20960 or play with the dataset yourself: huggingface.co/datasets/cop...

Work done with my amazing colleagues @aicentre.dk
@srishtiy.bsky.social @mariaa.bsky.social @serge.belongie.com @iaugenstein.bsky.social

Using our method, you can also get issue-specific frames inductively from the article texts. When publishers are compared across the political spectrum, some clear patterns of how the left frames Immigration vs the right.

Across topics, we find substantial differences in framing across the article and the image. These hold across political leanings as well.

We collect a dataset of 500k articles and images from various publishers in the US, across the political spectrum and systematically analyse differences in framing across them.

Editors choose to convey more subtle messaging through images that can evoke a more emotional response. But can this be measured? We demonstrate a methodology using large language and vision models to do such multi-modal analysis reliably & at scale. We use both generic and issue-specific frames!