Link content using AI

[armantorkzaban]

Type of Issue

• [x] New Feature Implementation
• [x] Addition / Enhancement / Improvement

Type of Work

• [ ] Screen
• [ ] Widget/Component
• [ ] Business Logic (Service, Repository, BloC)
• [x] Backend Service
• [ ] DevOps
• [ ] Documentation

Description
Cross-reference entities based on rubric, context, relevance using AI

As a normal user,
I want to know who where else is talking about the similar or a related topic without extensive direct categorisation,
so that I can find and inform my network/like-minded actors and enter a collaboration faster.

---

(not used items) Actionable Tasks

• [ ] Implement a Cubit/BloC for [action] with states [...] and methods [...]
• [ ] Implement Widget to trigger [action]
• [ ] Implement Widget to display [result of the action]
• [ ] ...

Acceptance criteria

• [ ] Screen should render the initial UI properly
• [ ] After [action is performed] state should change
• [ ] New UI is properly rendered
• [ ] ...

[lukas-h]

This could more easily be implemented with tags, where every person on the platform sets their interests, every post is referenced with tags. This would allow for some very simple algos, possibly without even using any higher AI models

[armantorkzaban]

@zanzendegi WDYT?

[IranNewsWatch]

Self-proclaimed tags could be easily misused. If so, we need a tag-police which will be unsustainable. Yet I agree that starting with tags could be a good start. It can give us some categorized data (assuming some reliability) to build our AI around it. This method is dumber than the AI model but easier to code. However, we should keep in mind that clustering by the similarity of content is different from categorizing (assigning to defined categories). The latter is more complicated and needs a complex NLP system.