Lea Frermann retweeted
1/ Caregivers teach infants how to speak by correcting errors.
Child: "I goed to the park"
Caregiver: "I went to the park"
Can models do this too? Read our recent paper on reformulation in image captioning:
arxiv.org/abs/2501.04513
w\
@AbendOmri
@leafrermann
@GabiStanovsky
Lea Frermann retweeted
This work was done under the supervision of @leafrermann and @eltimster: 📰 #MediaBias might be subtle, but it leaves its mark in the #Framing of an article. Hence looking at both is of major relevance. In our survey, we found three main characteristics in Framing: 👇🏽
Lea Frermann retweeted
We are building a National Climate Policy Database with @MCFunimelb - calling on Australian businesses, organisations, and governments to submit their climate plans and reports: bit.ly/3s9XXgN
#ClimateAction #policy #datascience #datacollection
pursuit.unimelb.edu.au/artic…
Lea Frermann retweeted
Happy to share that our paper "Screenplay Summarization Using Latent Narrative Structure" w/ @mlapata, @frank_e_keller and @leafrermann is accepted as a long paper to #acl2020nlp. We identify an optimal sequence of scenes and construct video summaries of CSI episodes. Example:
Lea Frermann retweeted
We're hiring: one 2-year postdoc position (go.unimelb.edu.au/c4gr) and 2 fully-funded PhDs (go.unimelb.edu.au/4tgr) to work on project with Trevor Cohn, @leafrermann, and me on fairness for #NLProc
Lea Frermann retweeted
More press coverage for Frermann et al's paper on solving crime drama with NLP: theregister.co.uk/2017/11/02…







