"Taxonia", an iNaturalist-based taxonomy quiz

Hello, iNaturalist users! I’m a full-stack/software developer who has been using iNaturalist on the side for a couple of years now. I decided to put my interests together and cook up a website that would allow users to practice identifying organisms (in particular, their taxonomic classifications; e.g. Latin names). I’ve searched and found a couple alternatives that have been developed before, but my intent with this is to solve a couple problems at once that I felt weren’t addressed as well as they could be:

  1. Helping people learn to identify the organisms
  2. Helping people learn the Latin designations, so over time you can better understand the actual taxonomy.

I’ve deployed my app at taxonia.app

The code is open source. For bug reports, feature suggestions, or to keep tabs on changing I’ll be making, the best place is the GitHub repo.

I’ll also be receptive to discussion here!

31 Likes

There was one photo of a bunch of ring-necked ducks and one canvasback in the background, I put ring-necked duck, but the answer was canvasback. I don’t know how to fix this annoying issue that taxon quizzes often have. Also, I ID-ed most of them correctly, but I got only one “correct” when I put the Latin name in and not the English common name. Is this only for learning Latin names?

3 Likes

Yep, right now it’s aimed at learning the Latin names. There’s a few ways to approach this problem I’ve been thinking of

  1. An option that targets either the Latin name or the common name (although it should be understood that common names might not be widely agreed on, in many cases).
  • I would also note that this doesn’t makes as much sense if, in the future, I add the option to test for genus/family names, so it would be specific to species or subspecies names
  1. If the photo has multiple subjects, distinguishing them might not be an easy problem to solve in every case, but there could be some approaches that work most of the time. For example, if the test is based on a Kingdom and you’re aiming to identify the Species, I could add a hint system to show you the family or even maybe the genus.
2 Likes

I wanted to try it out, but since it’s only about American species I didn’t even start. I live in Europe.

6 Likes

Those are only example tests since I’m in North America; you can use the New Test button to open a modal and select a completely different region (like Europe, or a specific country, or even counties/provinces, etc)

7 Likes

Amazing work! This is really well done!

4 Likes

I tested it out, and it was pretty fun! It was simple but nice, although the fact that I could only type scientific names tripped me up a bit. But overall, this is awesome! Please continue improving and adding to this!

1 Like

This reminds me a lot of https://identifi.life/, which is basically the same thing but with more levels.

1 Like

Scientific names use both Latin AND Greek.
This taxon uses both Greek for chrys and Latin for aurea

1 Like

Nice! Just the accuracy calculation seems hard to interpret: if I have correct genus name and the species name is wrong, I can have score 0. Only apparently when there are at least few letters in my guessed species name overlapping with the correct name, the score will be > 0. I’d suggest to modify the scoring function - it is likely not important but I think users will enjoy more to see non-zero scores when they actually made potentially a very good guess and have genus name correct. What about normalizing the score so the correct genus name guarants 0.5points, species name another 0.5 points. I’d be anyways curious to see the scoring function: is for example first letter of name guessed correctly bringing higher score than e.g. fifth letter? I think such scoring would naturally reflect the way we remember names. I’d love to also see the full taxonomic hierarchy for each photo in the final quiz evaluation - I think that way users would learn a lot. Also consider increasing diversity of the quiz: e.g. avoid displaying same species multiple times.

3 Likes

Thank you! I do plan to continue adding to this and making it more feature-rich and effective

Good catch, I’ll be sure to keep that in mind in communication, and to be clear in documentation what the app is expecting of users

1 Like

Right now I’m using a rudimentary string similarity algorithm (Levenshtein distance) with a little bit of a curve (<= 50% = 0, >= 95% = 1.0). It’s a little arbitrary, but just what I landed on after practicing with it for a little bit. These are good suggestions, I’ll be sure to add score improvement/transparency as an issue to track, as well as the rank hierarchy idea!

This is fun! Thanks for putting it together. I notice accuracy requires proper punctuation (nice).

Guess: salvia farinacea
Actual: Salvia farinacea
Common name: Mealy Blue Sage
Accuracy: 0.972

Sometimes I’ll type in Latin names when running through the ID queue but this has made me realize on many of these I’m relying on knowledge of the first four or five letters to get the suggestion.

2 Likes

This is so great! Thanks for making this community resource.I was actually just thinking about this this morning, wondering if it existed. Couple places to go next:

-Ability to downvote photos that just aren’t clear enough for reasonable ID
-Filter that takes out duplicates (I made a grasses of humboldt county list, Briza maxima showed up like 8/15 times)

I also really liked that you get a partial score for getting close with a Latin name. Very cool!

7 Likes

This is amazing and quite fun! would it be possible to make one for California native plants? It would be good to help my friends learn about native plants. These type of tests are really good for memorizing

1 Like

Fun! My ability to ID was better than I thought it would be. I guess it goes to show that I usually end up being pretty conservative with my IDs, or end up puzzling over tricky hybrids and things of that nature.

I’d love to see more options in line with the iNaturalist filters. It would be particularly great to have an option to ID the cultivated plants of a location–it seems to be only wild as of now.

2 Likes

Well made.
I had a lot of fun trying to guess the espices using latin names- especially when the photo was mine. There is just one small problem with the locations: I selected to play in the brazilian state of Rio Grande do Norte, but, sometimes, appears photos from the Fernando de Noronha island, which belongs to another state.

1 Like

Nice! I was also working on something very similar, but I lack motivation to continue working on it, you can try it here if you want to borrow some ideas : https://speciesguessr.univ-lyon1.fr/

2 Likes

Looks like the place information is drawn from iNaturalist, so the boundaries of your state and the correct one will need to be amended here.

1 Like