An image of a robot struggling with numbreed blocks. AI has a problem with random ordering.

Totally Random! Getting Round AI Random Blindness in Worksheet Creation

If you’re already using AI for language learning content creation, you’ve probably already cried in horror at one of its biggest limitations. It’s terrible at putting items in a random order.

Random order in language learning exercises is pretty essential. For instance, a ‘missing words’ key below a gap-fill exercise should never list words in the same order as the questions they belong to.

Obvious, right? Well, to AI, it isn’t!

Just take the following prompt, which creates a mini worksheet with an introductory text and a related gap-fill exercise:

I am learning French, and you are a language teacher and content creator, highly skilled in worksheet creation.
Create a French worksheet for me on the topic “Environmentally-Friendly Travel”. The language level should be A2 on the CEFR scale, with clear language and a range of vocabulary and constructions.
The worksheet starts with a short text in the target language (around 250 words) introducing the topic.
Then, there follows a gap-fill exercise; this consists of ten sentences on the topic, related to the introductory text. A key content word is removed from each sentence for the student to fill in. For instance, ‘je —— en train’ (where ‘voyage’ is removed).
Give a list of the removed words in a random order below the exercise.

The output is very hit and miss – and much more miss! Perhaps 90% of the time, ChatGPT lists the answer key in the order of the questions. Either that, or it will produce feeble jumbling attempts, like reversing just the first two items on the list.

AI’s Random Issue

One prompt-tweaking tip you can try in these cases is SHOUTING. Writing this instruction in caps can sometimes increase the bullseyes. Put them IN RANDOM ORDER, darn it! It doesn’t help much here, though. It just doesn’t seem worth relying on Large Language Models like ChatGPT to produce random results.

The reason has something to do with the fundamental way these platforms function. They’re probability machines, guessing what word should come next based on calculations of how likely word X, Y or Z will be next. Their whole rationale is not to be random; you might even call then anti-random machines.

No wonder they’re rubbish at it!

A Road Less Random

So how can we get round this in a reliable way that works every time?

The simplest fix, I’ve found, is to find another, non-random way to list things differently from the question order. And the easiest way to do that is to simply list things alphabetically:

I am learning French, and you are a language teacher and content creator, highly skilled in worksheet creation.
Create a French worksheet for me on the topic “Environmentally-Friendly Travel”. The language level should be A2 on the CEFR scale, with clear language and a range of vocabulary and constructions.
The worksheet starts with a short text in the target language (around 250 words) introducing the topic.
Then, there follows a gap-fill exercise; this consists of ten sentences on the topic, related to the introductory text. A key content word is removed from each sentence for the student to fill in. For instance, ‘je —— en train’ (where ‘voyage’ is removed).
Give a list of the removed words in alphabetical order below the exercise.

The likelihood of this order being the same as the questions is minimal. Hilariously, AI still manages to mess this order up at times, adding the odd one or two out-of-place at the end of the list, as if it forgot what it was doing, realised, and quickly bunged them back in. But the technique works just fine for avoiding the order giving the answers away.

A simple fix that basically ditches randomness completely, yes. But sometimes, the simplest fixes are the best!

Random blindness is a good reminder that AI isn’t a magical fix-all for language learning content creation. But, with an awareness of its limitations, we can still achieve some great results with workarounds.

Parallel text style learning, like Assimil courses, can be a great way to improve your fluency.

DIY Assimil : Parallel Text Learning with ChatGPT

Assimil language learning books are hugely popular in our polyglot community. And for good reason – many of us learn really effectively with its parallel text method.

They’re especially userful when the base language is another of our stronger languages, adding an element of triangulation. I learned a heap of Greek vocabulary from the French edition Le Grec sans Peine, at the same time as strengthening my (ever slightly wobbly) French.

Now, Assimil is already available in a great range of language pairs. But it’s not always a perfect fit. For example, some editions are more up-to-date than others. More off-the-beaten-track languages still aren’t available. And at times, you can’t find the right base language – no use learning Breton through French, if you don’t have any French.

Enter ChatGPT (or your alternative LLM of choiceBing also does a great job of these!).

DIY Assimil Prompting

Copy and paste this into your AI chat, changing the language (top), translation language (middle) and topic (bottom) to suit.

You are an expert creator of language learning resources. I want to create some text-based learning units for beginner Malay learners (level A0/A1 on the CEFR scale). The units follow the parallel text approach of the well-known Assimil language learning books.

Each unit has a text in the target language (about 250 words) on a specific vocabulary topic. It should be narrative, talking about how the topic relates to an everyday person. It should be divided into logical paragraphs. After each paragraph, there is an English translation of that paragraph in italics.

The text should be written in very clear, simple language. The language must read like a native speaker wrote it, and be error-free and natural-sounding. Source the info for the text from target language resources online, making it as up-to-date and authentic as possible. It should be completely original and not copied or lifted from any other source directly.

After the text, there is a glossary list of the key topic words from the text, sorted alphabetically and grouped by parts of speech (nouns, verbs, adjectives, adverbs etc.).

Are you ready to create some content? The first topic is: Mobile Technology

This prompt creates a prose-based parallel text unit. However, if you prefer dialogue-style texts, simply change the second paragraph of the prompt:

Each unit has a humorous dialogue in the target language (about 20 lines) on a specific vocabulary topic. The dialogue should relate the topic to everyday speakers through colloquial, idiomatic language.

The prompt works a treat in both ChatGPT Plus (paid) and Microsoft Bing (free). I also got very useable results in the free version of ChatGPT and Claude 2. It works so well as the focus is purely on what LLMs do best: spooling off creative text.

How Do I Use Them?

So, with your shiny, new Assimil-style units spooled off, what do you do with them?

Personally, I like to copy and paste the output into the notes app on my phone. That way, they make nice potted units to browse through when I have some spare moments on the bus or train. They’re equally handy copy-pasted into PDF documents that you can annotate on your phone or tablet.

Parallel text for Malay language learning created by AI

Parallel text in Malay and English created by AI

In terms of real-world use, the self-contained, chatty texts typically created make perfect material for the islands approach to improving spoken fluency. Create some units in topics that are likely to come up in conversation. Then, spend some time memorising the phrases by heart. You’ll be able to draw on them whenever you need in real-life conversation.

Enjoy prompts like these? Check out my book AI for Language Learners, which lists even more fun ways to get results without paying hefty course book price tags!

AI for Language Learners by Rich West-Soley; ChatGPT, Bing and more for your languages study

AI for Language Learners – Book Now Available!

It was a labour of love that happily took up most of my summer, and it’s finally out! I’m very chuffed to announce that my book AI for Language Learners is available on all Amazon stores.

 

The book is the product of months of tweaking, prodding and experimenting with emerging AI chat platforms. If you’re a Polyglossic regular, you’ll have seen some of those nascent techniques appear on the blog as I’ve developed and used them in my own learning. The blog has been a bedding ground for those first book ideas, and I’m thankful to everyone who has followed along with my own AI journey.

What we’ve come to call AI are, strictly speaking, actually large language models (LLMs). These LLMs arise from billions of words of training material – truly staggering amounts of data. The resulting super-text machines are perfect matches for subjects that benefit from a creative flair with words, and as language learners, wordplay is our currency. The book contains over 50 rich prompts for getting the absolute most out of AI’s impressive capacity for it.

The process has been huge fun. Of course, that’s thanks largely to the often unintentional humour our non-sentient friends ChatGPT, Bing and others. I try to get this across in the book, which has its fair share of lighthearted moments.

I hope you have as many smiles trying the recipes out as I did putting them together!

AI for Language Learners is available on Amazon Kindle (UK £2.99, US $2.99) or in paperback (UK £7.99, US $7.99). Even better: if you’re a Kindle Unlimited member, you can download and read it as part of your subscription.

A vocabulary learning plan generated by AI (Bing chat).

AI Prompts for Language Lesson Prep

I’ve talked previously about the importance of doing prep work for your iTalki and other one-to-one language lessons. Well, AI platforms just make that a lot easier (and free!).

You can generate mini ‘schemes of work’ that are matched precisely to the topics you are studying. With the proliferation of big name, public access AI platforms like Bing and ChatGPT, it’s easier than ever to support your learning in these incredibly tailored ways.

Sample Prompt

The best way to see just how useful this is, is to dive right in. Here’s a ready-made prompt to try with your chatbot of choice; simply change the language, topic and level to suit:

In a week’s time, I have an Italian conversation lesson on the topic of “Renewable Energy”. Could you please create some materials for me to prepare for it? I am level A2 in the language, so the language level of the materials should match that. I’d like:

– a vocabulary of up to 20 words and phrases relevant to the topic, with sample sentences (nouns, verbs, adjectives)
– some questions and answers modelling opinions on the topic using the vocabulary
– a brief introductory paragraph on the topic, again using the vocabulary
– a learning plan for the week
 

Tip: if your AI stops part of the way through (these outputs can be long!), just type Please continue and it will carry on spooling.

The final output will be a mini workbook-style guide to the topic, with vocab, discourse modelling, and reading practice. I like to paste mine into a text document, then export it as a PDF to read on my phone or tablet when I’m out and about. Here’s that sample Italian one (truncated to 10 items for brevity!):

Italian – Renewable Energy Topic Plan (PDF)

Learning instantly organised, in just a few clicks!

It’s All About You…

Experiment with that sample prompt to get exactly the kind of output that works specifically for you. For instance, you could tailor the output much more closely to your own circumstances by adding something like:

Please relate the material to recent developments in X/Y/Z.
 

…where X/Y/Z is your town, your field, your related interests. The sky really is the limit.

And a note of advice: when you find a prompt that works like a charm, save it in a note somewhere. It could become your go-to resource for future language learning.

Happy prompting!