Empowering journalism-approaches to automatic text summarization

on YouTube ↗22m

" summary

Sees 70% of words go unread and claims ~28% of content drives understanding and attention averages 8 seconds, so TX Services gathers article–summary pairs, trials Tool A and Tool B (market extractive systems and GPT‑3/DaVinci prompts), then trains a custom model with journalist-in-the-loop and a lightweight UI. Results show high grammaticality, improved fact capture and journalistic text quality, but low teaser usability; overall user experience is fast and workable. Ongoing work with Media Technology Center and ETH.

AI-written summary · about

" chapters (10)

1. 0:00Introduction
2. 1:06Attention in digital journalism
3. 4:15Journalism and AI history
4. 8:07NLP industry impact
5. 10:15Automatic news summarization
6. 12:42Evaluating summarization tools
7. 15:10Custom summarization model
8. 17:36Journalist feedback integration
9. 18:40Future research and conclusion
10. 19:45Q&A

" related talks

• 58mAutomatic text summarizationMasa Nekic · NDC Oslo 2019
• 19mPhrazIt : Tool for automatic text summarizationNiyati Parameswaran · Data By the Bay 2016
• 30mChallenges & Solutions for AI-Based Message SummarizationSwapnil Hemant Thorat · Conf42 Machine Learning 2025
• 07mScreen2Words: Automatic Mobile UI Summarization with Multimodal LearningYang Li, Gang Li, Tovi Grossman, Bryan Wang, Xin Zhou, Zhourong Chen · UIST 2021
• 36mSAMENVATTR, or, How to Make a Text Summarization Tool for your Language of ChoiceRobert Rodger · PyData Amsterdam 2018
• 29mAI in Media: Balancing Innovation, Quality, and ResponsibilityDominic Herzog · DSC Europe 23
• 26mUnsupervised learning for news summarisationJakub Kubajek · PyData Warsaw 2019
• 52mTowards fast and coherent long-form text generationMohit Iyyer · Data Science fwdays'20 online