Talk:Stance detection

Adding sections

Hello all! I have just created this page and plan on adding a section on Applications, as well as Multimodal and Multilingual stance detection. Jacob Rampino (talk) 17:17, 29 June 2026 (UTC)

This article was directly worked on and copied from my sandbox, as are the planned sections if you would like to review what I am working on. Jacob Rampino (talk) 20:05, 29 June 2026 (UTC)

Peer review

Stance detection

Hello! I recently created a Stance detection article and wanted a general peer review. It has largely been developed in a sandbox up until now. I am still working on sections regarding applications, as well as multimodal and multilingual models, so I am mostly interested in how the existing content or sourcing can be improved. I know it uses a few pre-prints, but I tried to keep those relatively limited to where they were relevant and with generally established authors. I look forward to any critiques and feedback! Jacob Rampino (talk) 18:02, 29 June 2026 (UTC)

Hey! Noob editor here, just trying to be helpful. I like this article! Just a few reactions.

• I really like intros that start with a barebones definition then expand. Focus on how it's an NLP task; I think you could take out the reference to computational linguistics, and move essential techniques (content analysis and text mining) to later in the paragraph.
• That survey paper (ALDayel) is a really strong reference. I appreciated the history bit.
• The header Supervised classification could probably be replaced with something nodding to the fact that the section is about traditional ML models. The task at hand could be called a supervised classification task, so this isn't super clear to me reading the subtitle what the following content will be.
• For the Deep Learning section it feels like we're really focused on the history of the task. I think that's fine for ordering, but within paragraphs, since it's a technical article, it might be better to start with the what. When skimming, people like to start a paragraph to see if they want to read the rest, so a what/why/how structure can be better than the why/what/how structure the paragraphs use right now. A little more nitpicky, I think talking about BERT you could frame how the papers specifically take advantage of the model structure structure and fine tuning techniques to do the task; right now a lot of the detail about BERT and fine-tuning feel like definitions I could get, more or less, by clicking the article names. I think the level of detail is fine, it just needs to feel justified.
• "generative models often cannot be archived for replication" is more of a researcher complaint than a fundamental flaw from a user perspective, so it might belong in a separate sentence.
• I do like that the article is framed as an NLP task, but with the nods to real-world use-case drawbacks in the LLM section, I would be interested if you can find sources in on actual real-world use. If so, that would be really interesting info!
• The Espinosa source is a decent experiment, but you kind of make it sound like a survey result. Not sure it really backs up something as strong as "no single model has been found to consistently dominate across tasks." It might be fine.

Hamolton (talk) 01:38, 5 August 2026 (UTC)

Thank you for this feedback! This is all incredibly actionable and I agree. I will make these changes. Jacob Rampino (talk) 02:17, 5 August 2026 (UTC)

Yay glad to be helpful! Hamolton (talk) 19:28, 5 August 2026 (UTC)

GuineaPigC77

This article looks well researched. Two things I noticed...

• Some of the sources look like preprints, which are considered unreliable Wikipedia:Reliable_sources/Perennial_sources#Preprints
• Per Wikipedia:Manual_of_Style/Lead_section#Citations I think the lead should only summarize what is alread cited later in the article.

I hope you find these short comments helpful! GuineaPig_C77 (𒅗𒌤) ☕ 02:17, 14 August 2026 (UTC)

Hi! I will definitely incorporate your comments for the lead. However, preprints are simply treated like SPS, and can be used reliably if the authors are subject-matter experts! Jacob Rampino (talk) 03:52, 14 August 2026 (UTC)