using Canvas to teach foreign languages that use accented letters and other special characters

I am currently teaching a high school Spanish course, and many of my assignments use the fill-in-the-blank question type in new quizzes. The problem that I'm encountering is when a question requires a special character, such as an accented vowel (examples include: á Ó or ü) or the special letter ñ (which is a different letter from N).

I cannot use the 'contains' option, because sometimes the correct answer will be a shortened version of a valid but incorrect answer. For example, habla means 'he speaks,' but hablas means 'you speak.' If a student were to write hablas where they were supposed to write only habla, it would be counted correct using the 'contains' option. For the same reason, I can't use the 'Close Enough' option; a difference of a single letter must be counted wrong (not just adding extra letters, but habla and hablo are two different valid words in Spanish, but one is a correct answer and the other is not; the 'Close Enough' option would mark the incorrect answer as correct).

The problem with the 'Exact match' answer is that it is case sensitive. So if I make the correct answer habla but the student capitalizes the word, Habla, it will be counted wrong.

I used the 'Specify correct answers' option for a while, but it gets tedious having to type out every possible capitalization for each question (especially in cases where a student might include the definite article; even having only all capitals, no capitals, and first letter capital, this means a bunch of different possible combinations: gato, Gato, GATO, el gato, El gato, El Gato, EL GATO, el Gato)

So I was delighted to discover the 'Regular Expression Match' option. I spent a lot of time making sure I had the RegEx answers correct so that it would accept any variation of the correct answers (with or without the definite article, regardless of capitalization). I was so excited!

Imagine my disappointment when the system counted wrong any correct answer that included an accented vowel or the ñ. I'm going to have to go through every student's test to manually fix the grading, which is exactly the opposite of what I wanted to use the quizzes for in the first place.

Is there any remedy for me? How do other language teachers handle this issue? Or can Instructure simply add an checkbox so that we can make the 'Exact match' and 'Specify correct answers' options to be case insensitive?

Find more posts tagged with

Instructor

Developers

Accepted answers

All comments

Jeff_F

Hello @YorkistJ chances are you considered using exact match with amended instructions specifying that lowercase letters are to be used. i.e., no capital letters. Would that be an option?

YorkistJ

The problem with that is that I don't trust students to follow directions that closely. Even those that do read and follow the directions make mistakes; it's entirely possible that a student may capitalize the first letter without realising it, or perhaps have caps lock turned on and fail to notice. Honestly, I worry that someone might accidentally put an extra space at the end of their answer. Add to that the possibility of words that need to be capitalised, like proper nouns, and thus rendering a blanket instruction to never use capital letters moot.

jeremyfw

Regular Expression Match
This should be the elegant solution, but here’s the catch:
Canvas regex matching is applied after an internal text-processing step that inconsistently handles Unicode. Accented characters often fail to match even when the regex itself is correct. This is a known limitation/bug, not user error.

So: what can you do today?

1. Use regex + character classes (partial mitigation)
You can sometimes get better results by explicitly enumerating accented variants:

^(el\s+)?h[aá]bl[aá]$

The winning strategy: Accent-aware character classes + word anchoring

Instead of telling Canvas “ignore accents,” you tell it exactly which accents are acceptable for this word and nowhere else.

Think in three layers:

Anchor the word (no contains leakage)
Control capitalization (without enumerating variants)
Explicitly model accents only where they belong

1. Case-insensitive without duplication

Canvas regex doesn’t support flags like /i, but you can simulate it safely:

[hH][aA][bB][lL][aA]

Yes, it’s ugly. But it’s deterministic and safe.

For longer words, you usually only need to do this for the first letter:

[Hh]abla

That already eliminates most capitalization false negatives.

2. Accent-aware vowels (the core technique)

Define localized vowel classes, only where accents are valid.

Spanish vowels:

a → [aá]
e → [eé]
i → [ií]
o → [oó]
u → [uúü] (context-dependent, but often acceptable)

So:

[Hh][aá]bl[aá]

This accepts:

habla
Habla
hábla (rare but sometimes typed)
Hábla

But it does not accept:

hablas
hablo
hablar

Because of anchoring (next step).

3. Word boundaries: no semantic leakage

Always anchor the answer:

^([Ee]l\s+)?[Hh][aá]bl[aá]$

What this does:

^ and $ block extra characters
Optional article allowed
Space required after article
No suffixes sneak in

This is the single most important anti-leakage rule.

4. Ñ must be explicit (never normalize this)

Ñ is not a variant of N. Treat it as sacred.

Correct:

[aA]ñ[oO]

Never do:

[nñ]

That creates catastrophic semantic leakage (ano vs año).

5. Reusable templates (this is where you save time)

You can pre-define mental templates:

Simple noun

^[Gg][aá]t[oó]$

Noun with article

^([Ee]l\s+)?[Gg][aá]t[oó]$

Verb (present, 3rd person)

^[Hh][aá]bl[aá]$

Verb with object pronoun

^[Dd][áa]m[eé]l[oó]$

You swap only the stem, not the structure.

What you should not do

Do not accept unaccented forms universally if accent matters instructionally
Do not use .*, \w, or contains
Do not normalize ñ with n
Do not rely on “Close Enough” for Romance languages

The honest pedagogical boundary

Canvas regex can help you reach 90–95% correctness safely.

The remaining 5–10% - especially:

stress-distinguishing accents
homographs
clitic clusters

… are better assessed in writing tasks or manual review, not auto-grading. That’s not a failure of regex; that’s respecting how Spanish encodes meaning.

YorkistJ

The problem I'm facing here is that it matters whether a word has an accent. For example, hablo and habló are two completely different words. The first means I speak, whilst the second means he spoke. So the [oó] option will not work.

What I've been doing is the [Hh][Aa][Bb][Ll][Oo] solution that you described above. It works perfectly for any words that are not accented. However, when I put accented characters in the mix, such as [Hh][Aa][Bb][Ll][Óó], Canvas counts the correctly accented answer (habló) wrong, but the incorrectly unaccented answer (hablo) correct.

Additionally, I found out after some experimentation that it counts accented letters as non-word characters. At first, I was using the /b tag delineate words, but I noticed that it treated the ó as a non-word character and thus incorrectly matched the word boundary. I don't know if this adds to the problem, or is a separate issue.

jeremyfw

1. Why `[Óó]` behaves “backwards” in Canvas

In proper Ruby 2.5 (Onigmo), this pattern:

[OoÓó]

would match either o or ó. There is nothing ambiguous about that.

Canvas, however, is doing something before the regex runs:

It appears to strip or normalize accented characters inconsistently
Then applies the regex to the post-processed string
But still compares against the original student input when determining correctness

The result is perverse but explainable:

habló → internally mishandled → fails literal comparison
hablo → survives normalization → matches regex → marked correct

So the system isn’t “misreading your regex.”
It’s losing the accent before the regex even gets a fair shot.

That’s why the wrong answer passes and the right one fails.

2. Accented letters treated as non-word characters (`\b` issue)

This is a second, independent confirmation of the same root problem.

In Unicode-aware regex engines:

ó is a word character
\b boundaries behave correctly

In ASCII-only or partially configured engines:

Word characters = [A-Za-z0-9_]
Everything else = non-word

Canvas is clearly operating closer to the second model.

So when they say:

it treated the ó as a non-word character

That tells us conclusively:

Canvas regex word semantics are ASCII-only, not Unicode-aware.

This is not how Ruby normally behaves. This is Canvas constraining Ruby.

3. Why this kills any “accent-sensitive but regex-based” solution

Here’s the key, unavoidable conclusion:

If Canvas treats accented characters as non-word characters and/or normalizes them inconsistently, then it is impossible to use regex to distinguish hablo from habló reliably.

No amount of clever character classes can fix that. You can’t regex your way around a preprocessing bug.

This is why:

[oó] fails
\b fails
correct accented answers are rejected
incorrect unaccented answers are accepted

The grading engine literally cannot see the distinction you are trying to enforce.

4. What does still work (within limits)

Given these constraints, the only Canvas-safe options are:

A. Exact Match + manual capitalization variants

Painful, but it preserves semantic truth.

B. Regex only for:

capitalization tolerance
optional articles
spacing

…but not for accent discrimination when the accent changes meaning.

C. Split assessment responsibility

A very common (and defensible) practice:

Auto-graded quizzes test morphology, syntax, agreement
Accent meaning (habló vs hablo, él vs el) assessed via:
- short writing
- constructed response
- oral checks

This isn’t pedagogical compromise; it’s tool realism.

5. What this means institutionally (and why this matters)

This is no longer a “how do I write a better regex?” issue.

This is a Canvas internationalization defect:

It breaks Spanish
It breaks French
It breaks Vietnamese (where my English department teachers struggle)
It breaks Arabic transliteration
It breaks accessibility for screen-reader normalization

That’s the framing that gets traction.

6. The honest bottom line

You can say this with full confidence:

The regex patterns are valid
The linguistic reasoning is correct
The failures are caused by Canvas’s ASCII-biased regex preprocessing
There is no regex-only workaround that preserves semantic accent distinctions

That’s not pessimism. That’s engineering reality.

YorkistJ

In other words, the answer to my question 'how do other language teachers handle this issue?' is simply 'try to get their students to use only lowercase letters' and/or 'type a whole bunch of variations as "Specify Correct Answers" options.' That's very disappointing.

Is there any way we can convince Canvas to add case-insensitivity to their platform? Or am I asking too much here?

DoraDextre

When was this implemented?Does it affect the new quizzes or does it affect the classic quizzes as well??

Unanswered Questions

Dropped Scores
I have Canvas set to drop the lowest quiz score, however, my students can only barely tell that the grade text is slightly grey when dropped. Can it have a tag like the "missing" and "exempt" ones in the student gradebook?
Batch edit assignment due dates by section
My Canvas shell has three different sections that all meet at different times throughout the week. I need to be able to batch edit assignment due dates by section, but right now Canvas can't filter by section for batch edit. My only option is now to manually (tediously) edit due dates for each assignment individually.
Change Group Assignments to NOT save if no groups are selected
Morning! One of the more frustrating aspects of Canvas are groups and the issues that show up when copying courses that have them. But now we have yet another issue. Most recently we have had instances where people mark an assignment/discussion as a group assignment but then do not link the discussion or assignment to any…
Has anyone experienced the "Empty Course" page for Speedgrader, when there actually are students assigned and active on the assignment?
So, we have encountered a few assignments where the Speedgrader presents the teacher with the "Empty Course" page, even though the course is not empty. I have discovered two New Quizzes in the same course where this has occurred, despite multiple students being assigned (at Module level, which passes the permission down to…
Problems with grading since update
A teacher has expressed concerns with the recent update. Exact concern is as follows: "My Canvas has been acting glitchy since the last "update." For instance, when I grade essays the highlight and cursor tools do not always work. I have to try highlighting several times,or come back later to finish. It's told me I "ran…