Perplexity Tasks for Language Learners

AI techniques to support language learning are pretty well-known now. From structured conversation partners to resource creators, LLM platforms have been embraced by the polyglot community.

Like many of us, I dip in and out of them almost unthinkingly now. Often, I’ll snap in a page from a chapter I’m working on with my Greek teacher, and it’ll help me prepare ahead of a lesson. Sometimes, I’ll get it to reel off a list of useful phrases on a topic I’m studying. LLMs can make great worksheet creators, too. In many ways, it’s simply a very interactive reference tool, giving (mostly) reliable answers but with a big nod to context.

I’d been pretty dogged in my choice of platform, sticking for the most part with ChatGPT Plus. Claude and Gemini were also in the mix, alongside some fun running local models. But for the most part, I thought my tool choices were pretty settled.

But then I gave Perplexity a whirl.

Perplexity – Task Master

Perplexity isn’t an LLM in the sense that ChatGPT, Gemini and Claude are. It uses LLM technology. But it’s actually more of an intelligent, context-sensitive search tool, that uses natural language APIs to turbo-boost its web-hunting activities.

I’d clearly not found that prospect very exciting, as I’d not gone near it until now. But thanks to a bundled free upgrade, I got to try the premium tier of late. And one particular feature stands out as potentially transformative for my learning habits: Perplexity Tasks.

Tasks are scheduled searches you set up with natural language instructions. And those instructions can be as rich as your usual LLM prompts in terms of requested formatting and such like, so in essence, you can build regular bulletins with up-to-date information in any language you like. Take one of mine, that runs daily:

Search the global news for the biggest world news story of the day. Summarise it in French, German, Modern Greek, Polish, Scottish Gaelic and Swahili at a level appropriate for an intermediate learner, ensuring that the translation is of the highest, native speaker standard quality, idiomatic and natural-sounding. Summaries should be 3-4 sentences long. Highlight key words in bold.

Accompany each summary text with a glossary / vocabulary list detailing all the key / difficult words from it in dictionary format (listing word class, irregular parts if applicable etc.). Hyperlink glossary items to Wiktionary entries where available with further information on them (use the English version en.wiktionary.com).

Lay it all out neatly to make it easy on the eye. Use plenty of emojis for impact too. Make this a fabulous resource for polyglot language learning! 🌍

Now, every morning, I get a wee news digest emailed straight to my inbox in multiple languages. It’s learner-friendly, includes vocab support, and gives me something to talk about in my language meets and lessons. I’ve done the same for academic paper searches in linguistics, and stories on dialect appearing in news outlets.

It feels like a proper game changer!

Tasking on Other Platforms

Now, you don’t need Perplexity to do this – it’s just one of the most user-friendly ways I’ve found to do it. If you have ChatGPT,  check out Scheduled Tasks. In Gemini, Scheduled Actions will do the trick for Pro members. Copilot is in on the game too. Others will no doubt follow suit shortly – clearly, task scheduling is becoming one of those features AI platforms are expected to have.

What I like about Perplexity, though, is that its whole raison d’être is the search – it feels particularly suited to web-based tasks like news digests. It’s also quite nice to keep the separation between my everyday LLM ramblings, and my more structured, scheduled items (use it for a few weeks and you’ll have clogged your timeline up with dozens of chats!).

If you’ve been looking for a way to make AI genuinely work for your learning rather than distract from it, try setting up a task or two – you might just find it becomes part of your morning ritual as well.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.