AI-generated sports news.

Gannett Stops Using AI To Write Articles For Now Because They Were Hilariously Terrible https://www.techdirt.com/2023/09/01/gannett-stops-using-ai-to-write-articles-for-now-because-they-were-hilariously-terrible/

#NLP #NLProc #generativeAI #SportsJournalism

IJCNLP-AACL 2023 has accepted exciting tutorials!
Check out the list!
http://afnlp.org/conferences/ijcnlp2023/wp/sample-page/tutorial/accepted-tutorials/
#AACL2023 #NLProc

4/

[3] Dhananjay Ashok and Zachary C. Lipton. 2023. PromptNER: Prompting For Named Entity Recognition. http://arxiv.org/abs/2305.15444

[4] https://paperswithcode.com/sota/named-entity-recognition-ner-on-conll-2003

[5] https://paperswithcode.com/sota/named-entity-recognition-on-genia

#NLP #NLProc #KnowledgeGraph #Paper

3/

REFERENCES

[1] Previous posts on ICL and why it works: https://www.linkedin.com/posts/benjaminhan_llms-iclr2023-mastodon-activity-7093281970792632320-XE87, https://www.linkedin.com/posts/benjaminhan_generativeai-gpt4-llm-activity-7045542002947457024-bGPt/, https://www.linkedin.com/posts/benjaminhan_llms-gpt3-nlp-activity-7073726814170337280-EED5/

[2] https://news.ycombinator.com/item?id=37443921

#NLP #NLProc #KnowledgeGraph #Paper

2/ One particular recent paper in the discussion proposed an ICL approach named PromptNER [3], and reported SOTA results among its ICL peers on well-known datasets such as CoNLL 2003 and GENIA. Compared its results with the SOTA *fine-tuning* (FT) solutions, FT outperforms on CoNLL by 10+ points (94.6 vs. 83.48; screenshot 1) [4], and outperforms on GENIA too by 20+ points (80.8 vs. 58.44; screenshot 2) [5].

#NLProc #KnowledgeGraph #Paper

Pour compléter une équipe dans le cadre d’un projet de visualisation automatique d’environnements sémantiques appliqué au domaine de la recherche (CompaSciences), l’Institut des sciences sociales de l’Université de Lausanne #UNIL met au concours un poste de

Développeur·euse #NLProc et #visualisation de données

#Job #Lausanne

https://career5.successfactors.eu/career?career_ns=job_listing&company=universitdP&career_job_req_id=21196

We're hiring a Postdoc to work across NLP, Digital Humanities, and Semantic Web. https://www.rug.nl/about-ug/work-with-us/job-opportunities/?details=00347-02S000AC8P

Bring your creativity and skills to study the evolution of online fictional narratives https://golemlab.eu

@gronlp
#AcademicChatter
#DH #LinkedData #KnowledgeGraphs #nlproc @digitalhumanities @TedUnderwood @scott_bot @christof @fotis_jannidis @zoeleblanc @quinnanya @folgertk @melvinwevers @albertoacerbi @EvelynGius @DHd @dh_potsdam

Putting together my presentation for tomorrow's HumEval 2023 session of NLP reproducibility (https://humeval.github.io). I will also be heading next week to Prague for INLG 2023 (https://inlg2023.github.io) with my master student. Hopefully get to see folks who I haven't seen in some time in person!

#NLProc #NaturalLanguageGeneration #HumEval #INLG

Darmstadt is a hub for #AI research – from #NLProc to Intelligent Autonomous Systems – that attracts worldwide attention. If you are interested in what we have to offer, a new info website of @TU Darmstadt provides further information:

➡️ https://www.ki.tu-darmstadt.de/en

Darmstadt is a hub for #AI research – from #NLProc to Intelligent Autonomous Systems – that attracts worldwide attention. If you are interested in what we have to offer, a new info website of @TU Darmstadt provides further information:

➡️ https://ki.tu-darmstadt.de/en

One week left until events kick off for #SIGDIALxINLG2023! You can still register to join us on-site in Prague or, if you can't make it all the way to Czechia, join us virtually!

Virtual registration is free, to enable as many people to join the conference as possible, so please share far and wide: https://sigdialinlg2023.github.io/

#PleaseBoost #NaturalLanguageGeneration #Dialogue #Discourse #NLProc #ComputationalLinguistics

Of the 3 papers I co-reviewed for #emnlp2023 , 2 used chatgpt for generation or evaluation.

Does anyone see a problem with littering our research with proprietary, black-box baselines?

#nlproc #nlp #OpenScience

“This statement is endorsed by the following members of the GPA’s International Enforcement Cooperation Working Group (“IEWG”).”

Joint statement on data scraping and the protection of privacy https://ico.org.uk/media/about-the-ico/documents/4026232/joint-statement-data-scraping-202308.pdf

#copyright #privacy #GenerativeAI #law #NLProc

11/end

[3] Previously: “Model Editing: Performing Digital Brain Surgery”. https://www.linkedin.com/posts/benjaminhan_llms-causal-papers-activity-7101756262576525313-bIge

[4] Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang, and Xindong Wu. 2023. Unifying Large Language Models and Knowledge Graphs: A Roadmap. http://arxiv.org/abs/2306.08302

#KnowledgeGraphs #GenerativeAI #LLMs #NLP #NLProc #Paper

8/ There are surely other benefits of using KGs to collect and organize knowledge. They do not require costly retraining to update, therefore can be updated more frequently to remove obsolete or incorrect facts. They allow more trackable reasoning and can offer better explanations. They make fact editing more straightforward and accountable (think of GDPR) compared to model editing [3].

#KnowledgeGraphs #GenerativeAI #LLMs #NLP #NLProc #Paper

7/ Their result shows that even #GPT4 achieves only 23.7% hit@1 on average, even when it scores up to 50% precision@1 using the earlier proposed LAMA benchmark (screenshot). Interestingly, smaller models like BERT can outperform GPT4 on bidirectional, compositional, and ambiguity benchmarks, indicating bigger is not necessarily better.

#KnowledgeGraphs #GenerativeAI #LLMs #NLP #NLProc #Paper

6/ In each benchmark, instead of asking LLMs to retrieve masked words from a cloze statement, it also asks the LLMs to retrieve all of the implied facts and compute scores accordingly (screenshot).

#KnowledgeGraphs #GenerativeAI #LLMs #NLP #NLProc #Paper

5/ Instead, they focus on more intricate topological and semantic attributes of facts, and propose 9 benchmarks testing modern LLMs’ capability in retrieving facts with the following attributes: symmetry, asymmetry, hierarchy, bidirectionality, compositionality, paths, entity-centricity, bias and ambiguity (screenshots).

#KnowledgeGraphs #GenerativeAI #LLMs #NLP #NLProc #Paper

3/ The result shows that even without specialized training, language models such as BERT-large can already retrieve decent amount of facts from their weights (screenshot).

#KnowledgeGraphs #GenerativeAI #LLMs #NLP #NLProc #Paper

2/ An early paper in 2019 [1] posited that compared to #KnowledgeGraphs, it is easier for language models to adapt to new data without human supervision, and they allow users to query about an open class of relations without much restriction. To measure the knowledge encoding capability, the authors construct the LAMA (Language Model Analysis) probe where facts are turned into cloze statements and language models are asked to predict the masked words (screenshot).

#NLP #NLProc #Paper