These AI projects have been loads of fun. I’ve been brainstorming about taking the project further and testing the bot’s limits. ChatGPT has proven somewhat helpful in streamlining tasks like a personal digital assistant. As an educator, I wondered how using ChatGPT can simplify lesson planning. Can it create Mad Libs assignments for me so I don’t have to myself?
Human fill-in-the-blanks (Mad Libs)
Fill-in-the-blank is a structure that has been widely used in composition exercises and helps students develop their grammar and vocabulary recall. There are dozens of ways to use the fill-in-the-blank format, but one thing remains constant: it’s an incredibly effective pedagogical tool. How can we use this structure in our ancient language classroom? There are many benefits to using composition exercises, including:
- Lacuna: Great way to parallel thinking about how texts survive and demonstrating how scholars make educated guesses!
- Composition: Students need to have decent grammar skills in order to complete Mad Libs
- It’s funny!
This article isn’t really focusing on if Mad Libs are good pedagogical tools or not, (I obviously am pro-Mad Libs,) but moreso interested in easily, quickly, and accurately creating these assignments using AI. There are some good resources about using Mad Libs in the Latin classroom like Latina Hilara’s post here, or existing Mad Libs resources like Magistrula.
Why Mad Libs?
Well, Mad Libs is recognizable to students. They might not understand what you mean when you ask them to parse a verb, but it’s pretty obvious what you want when you tell them, “Give me an adverb that ends in -ly!”
Mad Libs is an inherently flexible format, so you can introduce grammatical constructs in English using Mad Libs. Work your way up in complexity by having students complete Mad Libs in the target language.
They also have an educator’s guide on their website, which I’ve linked and attached here.
How to get ChatGPT to crank Mad Libs out for you
While these assignments might seem entertaining, creating Mad Libs can be time-consuming and annoying. Of course, you can do these by hand, and you’re (hopefully) faster than your students at analyzing words, but maybe you can spend that time grading work that they turned in last semester instead! So sit back, and let ChatGPT take the reins. The best part about using this method is increasing or decreasing how much the AI interferes with the text.
English method:
With this example, I just wanted to see if ChatGPT could make something into a Mad Libs. I wasn’t testing its grammar analyzing skills yet, so where there would be a blank space, I just asked it to write “[INSERT].”
This method is perfect for teachers who still control how the grammar in the blank spaces is described, but they don’t want to do the work of randomizing which words to remove.
Note: I copy-pasted the translation from theoi.com
Input:
Can you remove random words from the copied passage and replace those words with “[INSERT]”, then produce a list of the words you removed?
Here’s the passage: Arms and the man I sing, who first from the coasts of Troy, exiled by fate, came to Italy and Lavine shores; much buffeted on sea and land by violence from above, through cruel Juno’s unforgiving wrath, and much enduring in war also, till he should build a city and bring his gods to Latium; whence came the Latin race, the lords of Alba, and the lofty walls of Rome.
Output:
Sure! I have replaced random words with “[INSERT]” in the passage. Here it is:
[INSERT] and the [INSERT] I sing, [INSERT] first from the [INSERT] of [INSERT], [INSERT] by [INSERT], came to [INSERT] and [INSERT] [INSERT]; [INSERT] buffeted on [INSERT] and [INSERT] by [INSERT] from [INSERT], through [INSERT] [INSERT]’s [INSERT] [INSERT], and [INSERT] enduring in [INSERT] also, till [INSERT] should build [INSERT] [INSERT] and bring [INSERT] [INSERT] to [INSERT]; [INSERT] came the [INSERT] race, the lords of [INSERT], and the [INSERT] walls of [INSERT].
And here’s the list of the words I removed:
- Arms (noun)
- the (article)
- man (noun)
- who (pronoun)
- coasts (noun)
- Troy (proper noun)
- exiled (verb)
- fate (noun)
- Italy (proper noun)
- Lavine (proper noun)
- shores (noun)
- much (adjective)
- sea (noun)
- land (noun)
- violence (noun)
- above (preposition)
- cruel (adjective)
- Juno’s (proper noun)
- unforgiving (adjective)
- wrath (noun)
- enduring (verb)
- war (noun)
- also (adverb)
- he (pronoun)
- city (noun)
- his (pronoun)
- gods (noun)
- Latium (proper noun)
- whence (adverb)
- Latin (adjective)
- lords (noun)
- Alba (proper noun)
- lofty (adjective)
- Rome (proper noun)
Let me know if there’s anything else I can assist you with!
The following example is a step up from just putting “[INSERT]” In this example, ChatGPT analyzes the grammar.
Input:
Can you create a Madlibs from the copied passage by removing random words and replacing them with their grammatical function, i.e., singular noun, adjective, proper noun, adverb ending in -ly?
Output:
Sure! Here’s a Madlibs version of the passage:
[Adjective] and the [noun] I sing, who first from the [plural noun] of [proper noun], [verb] by [noun], came to [proper noun] and [proper noun] [noun]; much [verb past tense] on [noun] and [noun] by [noun] from [adverb ending in -ly], through [adjective] [proper noun]’s [adjective] [noun], and much [verb ending in -ing] in [noun] also, till he should [verb] a [noun] and bring his [noun plural] to [proper noun]; whence came the [adjective] [noun] race, the [plural noun] of [proper noun], and the [adjective] walls of [proper noun].
Now, it’s your turn to fill in the missing words with appropriate parts of speech!
Latin method:
Even though it seems that ChatGPT has accurately parsed each word, I wouldn’t trust it completely. I would recommend using this to decrease the time it takes to do long passages but definitely go over its work.
Producing Mad Libs in Latin is the same process as in English. You can still ramp up in complexity here, but on the students’ end, forcing them to comprehend cases if it’s a noun, adjective, or any other combination of grammar.
Here’s an example of the Latin Mad Libs if you want to provide students with the gender, number, and case of nouns.
Input:
Can you create a Madlibs from the copied passage by removing random nouns and replacing them with their grammatical case, gender, and number? Then produce a list of the words you removed.
Here’s the passage: Arma virumque canō, Trōiae quī prīmus ab ōrīs
Ītaliam, fātō profugus, Lāvīniaque vēnit
lītora, multum ille et terrīs iactātus et altō
vī superum saevae memorem Iūnōnis ob īram;
multa quoque et bellō passus, dum conderet urbem,
inferretque deōs Latiō, genus unde Latīnum,
Albānīque patrēs, atque altae moenia Rōmae.
Output:
Certainly! Here’s the Madlibs version of the passage with nouns replaced by their grammatical case, gender, and number:
[Accusative plural noun] [nominative singular noun] canō, [nominative singular proper noun] quī prīmus ab ōrīs
Ītaliam, fātō profugus, [nominative singular proper noun] vēnit
[Accusative plural noun], multum [nominative singular pronoun] et [dative/ablative plural noun] iactātus et [ablative singular noun]
[vocative plural noun] superum saevae memorem [genitive singular proper noun] ob īram;
[Accusative plural noun] quoque et [ablative singular noun] passus, dum conderet [accusative singular noun],
inferretque deōs [dative/ablative singular proper noun], genus unde [nominative/accusative singular proper noun],
[Accusative plural proper noun] patrēs, atque [nominative singular adjective] moenia [genitive singular proper noun].
And here’s the list of the words I removed:
- Arma (nominative/accusative plural noun)
- virumque (nominative/accusative singular noun)
- Trōiae (nominative/accusative singular proper noun)
- quī (nominative/accusative singular pronoun)
- prīmus (nominative singular adjective)
- ab (ablative preposition)
- ōrīs (ablative plural noun)
- Ītaliam (accusative singular proper noun)
- fātō (ablative singular noun)
- profugus (nominative singular adjective)
- Lāvīniaque (nominative/accusative singular proper noun)
- vēnit (nominative/accusative singular verb)
- lītora (accusative plural noun)
- multum (nominative singular pronoun)
- ille (nominative singular pronoun)
- terrīs (dative/ablative plural noun)
- iactātus (ablative singular noun)
- altō (ablative singular noun)
- vī (ablative singular noun)
- superum (vocative plural noun)
- saevae (genitive singular adjective)
- memorem (genitive singular adjective)
- Iūnōnis (genitive singular proper noun)
- ob (accusative/ablative preposition)
- īram (ablative singular noun)
- multa (accusative plural noun)
- quoque (accusative plural noun)
- bellō (ablative singular noun)
- passus (nominative singular verb)
- dum (conjunction)
- conderet (accusative singular verb)
- urbem (accusative singular noun)
- inferretque (accusative singular verb)
- deōs (dative/ablative singular noun)
- Latiō (dative/ablative singular proper noun)
- genus (nominative/accusative singular noun)
- unde (ablative preposition)
- Latīnum (nominative/accusative singular proper noun)
- Albānīque (accusative plural proper noun)
- patrēs (nominative plural noun)
- altae (nominative singular adjective)
- moenia (genitive singular noun)
- Rōmae (genitive singular proper noun)
Let me know if there’s anything else I can assist you with!
It didn’t grab the gender, but it did a pretty good job otherwise!
Let’s try to make it harder on the AI by asking it to make Mad Libs with only verbs now.
Input:
Can you create a Madlibs from the copied passage by removing random verbs and replacing them with their grammatical person, number, tense, voice, and mood? Then produce a list of the words you removed.
Here’s the passage: Arma virumque canō, Trōiae quī prīmus ab ōrīs
Ītaliam, fātō profugus, Lāvīniaque vēnit
lītora, multum ille et terrīs iactātus et altō
vī superum saevae memorem Iūnōnis ob īram;
multa quoque et bellō passus, dum conderet urbem,
inferretque deōs Latiō, genus unde Latīnum,
Albānīque patrēs, atque altae moenia Rōmae.
Output:
Sure, here you go:
Arma virumque [INSERT] , Trōiae quī prīmus ab ōrīs
Ītaliam, fātō profugus, Lāvīniaque [INSERT]
lītora, multum ille et terrīs [INSERT] et altō
vī superum saevae memorem Iūnōnis [INSERT] īram;
multa quoque et bellō [INSERT] , dum conderet urbem,
inferretque deōs Latiō, genus unde Latīnum,
Albānīque patrēs, atque altae moenia Rōmae.
List of removed words:
- canō (1st person singular present active indicative)
- vēnit (3rd person singular perfect active indicative)
- iactātus (masculine singular nominative perfect passive participle)
- ob (preposition taking accusative case)
- passus (masculine singular nominative perfect passive participle)
- inferretque (3rd person singular present active subjunctive)
Conclusion: Mad Libs AI works!
Alright, well, I’m pretty happy with how ChatGPT did on this. It’s saved me a ton of time from creating these composition assignments by hand, and I can mess around with it even more to ramp up complexity based on my students’ skills.
If you want to check out my other AI experiments, you can find those here.
And if you liked this, be sure to subscribe to my newsletter to get updates on any other posts.