Dichotomous keys

[Silurus]

How easy/difficult would it be to incorporate dichotomous keys into the Cat-eLog?

[Jools]

Can you give me an online example of this in practice?

Jools

[Silurus]

http://tolweb.org/tree?group=Aspredinid ... luriformes

Instead of a key to genera, maybe a key to species?

[Jools]

I've looked into this and it'd be a huge amount of work and has a few technicalities that would require new features in the database.

Mainly this is because at arriving in a species via such a key, you often have intermediate steps between genus and species.

i.e.


1. Is fish yellow ................ species1

is fine but


1. Is fish yellow ................. 3
2. Is fish red .................... species1
3. Has fish got blue spots ........ species2
4. Has fish got no spots .......... species3

Is hard becuase you need an intermediate bucket, link, call it what you will for all the species linked from question 1 that could then be species 2 or 3. This is also the simplest example. Consider Corydoras and the mind boggles.

Also, many of the questions might be "Is the fish from xxx drainage" and thus renders the key useless in many "aquarist facing" cases.

I'd really like to do this, but the amount of work is staggering. Perhaps the first stage is either:

A key for the catfish families or
A key for a nice size of genus, around 20 species say.

Jools

[Jools]

HH,

Always liked this one, so.....

Borrowing some wiki-type code, I've now built a way to do this that is integrated with the cat-elog at any or all of species genus or family level. It might get a bit clunky with classification revisions happen, but that's a bridge to be crossed later.

Where do we start?

Jools

[housewren]

Jools,

I'm a rank novice/amateur when it comes to catfish, having only paid them much attention (only corys so far) for a little over a year now. But I have been trying to make myself a key for the genus Corydoras. I know, I am probably in way over my head! but I've been having fun working on it--it's as good as a 5000 piece jigsaw puzzle! [Still haven't tackled those spotted cories with a dorsal blotch partly on the body and partly on the dorsal fin yet (e.g., delphax, melanistius, ambiacus, agassizi, etc.)--now that should be a real challenge!]

Regarding location data being required to positively identify some species: I decided that the way to go would be to use drainage only as a last deciding factor, if required. That way the hobbyist could at least know that he has either species A or B. As a hobbyist I would be satisfied to know I have come as close as possible to identifying it without location information.

It would be nice to have a key to families, as a general starting point for someone trying to identify their fish; a key that uses only external characteristics that a hobbyist could observe without threatening the life of his fish.

Can this key be done piecemeal (e.g., someone writes a key for the species in one genus of loriicaridae, and later on someone writes a key to the genera of loriicaridae) or does it need to be done from the top down (e.g., key to the families first, then key to the genera within a family, and finally, keys to the species within the genera)?

Sounds like a fun project to me.

[Silurus]

Jools,

We could start with a small group for which a published key alreay exists (granted that most keys are not really hobbyist friendly). Pictures of all the species are ideal, but not entirely necessary.
Something like Rita would be good. A published key already exists (see the description of R. macracanthus), and we have pictures of most of the species.

[Jools]

But I have been trying to make myself a key for the genus Corydoras.

Well, that's something I'd been thinking about too. Where have you got with it? How many species?

It would be nice to have a key to families, as a general starting point for someone trying to identify their fish; a key that uses only external characteristics that a hobbyist could observe without threatening the life of his fish.

I'd really like to publish this too. I guess it should be put in as a new site suggestion, but maybe someone like Shane or Lee could be cajoled into working something out.

Can this key be done piecemeal (e.g., someone writes a key for the species in one genus of loriicaridae, and later on someone writes a key to the genera of loriicaridae)

It can be done in any order, but each unit tackled must be completed. The issue will be were we have species we wish to talk about but no entries for them in the cat-elog but that's for me to worry about.

Jools

[Jools]

A published key already exists (see the description of R. macracanthus), and we have pictures of most of the species.

Err, we don't have R. macracanthus in the cat-elog? Can you post the key here for starters?

Jools

[Silurus]

The key is in here.

[loachy_406]

Imo it would be very useful to have something like this to easily and quickly get you to a good starting place to IDing cats. A lot of the time I would like to help people, but I don't know where to start looking, so I can't without looking through every single cat. I would mostly be for more inexperienced people, but even experts don't (normally) know everything! Even if it only gets you to a genus, that is still a long way into the search. This new feature would take nearly forever to make, so it may not be possible in the near future, if ever. Jools has probably already thought of it and dismissed the idea. He seems to have thought of everything! Any thoughts or ideas?

[Marc van Arc]

You are certainly very enthousiastic, full of ideas and willing to help. That's okay.
The thing is that if you can't help someone you should accept that. People have little use for replies like: "I don't have one, but I like the looks of it" (or something to that extent. Btw this is NO quote of yours, but just an example).
What you could do is follow the post - most likely in What's my catfish - find out what it is eventually and learn from it. By doing so you'll get more and more experienced in identifying catfishes, thus being able to really help people with their questions. Everyone here has been through that stage, so don't worry: you're not alone

[Silurus]

This is an idea that has been previously discussed, but seems to have gotten nowhere.

[loachy_406]

Ok then. I'll just have to gather experience and become old and wise.

[MatsP]

It's not the most "hobbyist friendly", but it's a "key" in slightly less computerized model:
http://www.auburn.edu/academic/science_ ... y/key.html

--
Mats

[loachy_406]

I see what you mean! Thanks anyway.

[Bas Pels]

being a beginner in catfish myself, I often try to get the family of queries

Obviously, some families are easy, but others........

and in well known families I try to go further - without posting

[Jools]

I tried doing it just for families and it's not easy let me tell you!

The simple fact is that if someone can provide the information I can certainly put it online and once it's in the database all sorts of things become possible.

What I don't really have time to do is define all the rules, it's not just that but it's things like providing pictures and diagrams of the characteristics and talking around internal stuff that you can't show easily.

Bottom line, give me the info, families, genera and species and I will build it. I do not have the time (and likely won't for the foreseeable future) to research, write up and build this.

Jools

[MatsP]

I agree with Jools. The programming of it is fairly easy, in programming terms it's a simple binary tree structure where each question either leads to the left or right half of the tree, and eventually we'd get to a leaf-node: Success, we have our species.

The problem of it is definitely collecting, collating and formulating the data in such a way that it makes it work.

The starting point is already there: we have identifciation data fields for species and genus in the Cat-eLog. Unfortunatly, I don't believe even 5% of the species have any information in the species identification field. So finding (at least most of) the other 95% would be a task and a half.

I spent a large amount of time filling in the occurrence data. That was quite a lot of work, but most of the data is relatively easily accessible: if it's not in the Cat-eLog, it is in Fishbase/COF for most described fishes. L-numbers are also fairly well described in this aspect. So it's a few clicks of the mouse to get that data. Figiuring out how you differentiate two similar species of Cory is not so easy.

Further, some questions would definitely be of the form "Does it come from Brazil" or "Does it come from Rio Xingu" - because the location is essentially the only way to determine if the fish is species X or species Y.

--
Mats

[Jools]

Further to Mats comments, all undescribed fish would have to be included on external features alone too.

I'd suggest that it might be better to follow a non-scientific approach to this.

If we take the ultimate goal as a function that asks the user with an unknown catfish a question, they answer and depending on that answer another question is asked until they drill down to a single species - there is another way to get close to it.

That is to tag each species with attributes. Things like, suckermouth, white spots, blue eye, vertical bar through eye, feathery barbles, brown body etc. If all species are meticulously tagged, and they could have maybe even dozens of tags per species, you can drill down to at least a handful of species quickly and then put in some refining at the very bottom for things like number of soft dorsal rays etc. Thinking about it, building a "tag assigning facility" for the cat-elog team would not take more than a couple of days. The facility for end users to use it another similar time frame. Then it's up to the team...

Incidentally, I wrote a piece of software similar to this to identify the first 100 species of Corydoras way back when I needed to build an expert system to pass 2nd year at University. A lot smaller, but the same concept and certainly flagged up a lot of the issues.

Another point is that scientific descriptions will tell you perhaps how a species fits into its genus, or perhaps how it differs from congeners, if it's a good one then all congeners. I've yet to see a description paper that helps with family identification, it's always defined in the title. However Planet users will often not know the family and so are not off to a good start.

If there's an appetite for this approach, I am happy to consider it.

Jools

[MatsP]

Jools,

I like your thinking - and now that you mention it, I have been thinking along those lines myself (e.g. how can we "find all spotted Loricariidae"), although never got as far [in my thinking - never mind actually implementing it] as actually identifying with it.

--
Mats

[racoll]

A dichotomous key used by amateurs, with the number of taxa we are considering here, would have an unacceptable error rate.

However a matrix key with lots of photos and tags would probably work for the common/easy species, at least to generic level.

Check out a nice example here: http://www.lucidcentral.org/keys/aquari ... /Home.html

Although I'm not sure there is really a need for it. We already have a excellent identification tool called the "what is my catfish forum" which is more accurate than any computer based key and can operate on a variety of data sources!

[Marc van Arc]

Although I'm not sure there is really a need for it. We already have a excellent identification tool called the "what is my catfish forum" which is more accurate than any computer based key and can operate on a variety of data sources!

Hear, hear. That's the spirit!

[Jools]

I agree, however, wouldn't it be nice to have the easy ones filtered out? I've been thinking about this idea a lot recently and I think it has legs. Unfortunately I only have one pair of them and indeed hands and there are a few things they are needed for first. However, it occurs to me that collecting the data can happen before it is collated and used.

If I built the thing to allow you chaps to start tagging, would the team be up for the long haul?

Jools

[MatsP]

I agree, however, wouldn't it be nice to have the easy ones filtered out? I've been thinking about this idea a lot recently and I think it has legs. Unfortunately I only have one pair of them and indeed hands and there are a few things they are needed for first. However, it occurs to me that collecting the data can happen before it is collated and used.

If I built the thing to allow you chaps to start tagging, would the team be up for the long haul?

Jools

Sure, as long as the "long haul" is the usual Planet Catfish timescale: It will be ready when it's done...

--
Mats

[Martin S]

Sure, as long as the "long haul" is the usual Planet Catfish timescale: It will be ready when it's done...

--
Mats

If, in the same timescale, there is anything I can do to help, please ask!
Martin

[Jools]

** Just a reminder to talk about this when (some?) of us meet up at the catconv. **

Jools

[MatsP]

Continuing the "July bug blast" that got somewhat interrupted by other events for me...

Is this going somewhere?

I wrote a piece of code (in C++, but relatively trivial to port to PHP) that would generate the "intermediate steps" and links, from a human readable, comma-separated list of keys (with a few formatting rules - we could quite easily make an "editor" for this that can be used to enter it into the database so that it's much less manual work [obviously, there's still a "yes/no" type question to enter]).

Edit: Merged another thread into this one...

--
Mats

[Jools]

Is this going somewhere?

No. At least not in any kind of sensible timescale. That's to say I think we could produce a technical solution but I do not get the feeling that anyone would be willing to enter and manage all the data.

Less concrete, but I also get the feeling that keys will be overtaken by some technical advance or another over the next decade.

Unless any violent objection and in lieu of a squad of volunteers, I would moved to resolved as "discussed and not a go-er".

Jools

[MatsP]

I thought that would be the case... And we may want to look into Markos suggestion from the CSG conference...

--
Mats