Calling all Information Scientists

[Yuri_Rage]

For someone supposedly schooled in the art (or perhaps even science) of information, the OP is having an awfully hard time conveying the same. I would like to see an update, as an academic venture attempting to somehow quantify vague and/or subjective concepts (like sensory perception of beer) is interesting...if that is indeed the intent of the OP.

 

[blakelyc]

Well, this thread had gotten interesting A few items related to earlier posts:

Is there something more?

Yes... quantitative characteristics are necessary, but not sufficient to fully define a beer style. They get you in the neighborhood, though.

yet you have said literally nothing about which properties of beer you would like to ontologize. There are a handful of quantifications available, but the bulk of a beer's characteristics are highly subjective. This isn't a formal system.

This is the problem space, more or less. We have a set of measured criteria, but the natural language descriptions contain most of the real interesting information. If it was all measurement, then the problem of defining a beer style is closed-world, easy, and boring. But, as you point out, that isn't the case. The question of what properties of beer I want to ontologize is an interesting one, because I don't know. The point is to develop an ontology that attempts to model all the characteristics of a style. Parsing the language and defining the characteristics is very much part of the work.

The BJCP style guide provides a shared vocabulary for description, but beyond that your enthusiasm for classification seems to lack a clear understanding of why classification might be justified. Sometimes, when all you've got is a hammer...

Except they don't... nor does any other beer style guide. Style guides contain the descriptions themselves, not a shared vocabulary for making such descriptions. Aside from that, I have no great love for classification. That's a taxonomist's job, and I very much appreciate the work those folks do... but it ain't me. Ontological modeling is not about classification. It is about defining exactly what is means to be a member of a class, and what your relationships are to things of other classes.

For someone supposedly schooled in the art (or perhaps even science) of information, the OP is having an awfully hard time conveying the same. I would like to see an update, as an academic venture attempting to somehow quantify vague and/or subjective concepts (like sensory perception of beer) is interesting...if that is indeed the intent of the OP.

While I didn't intend for this thread to devolve into an alpha-dog discussion of one's resume and what one is supposedly schooled in (see the hammer comment, above), I am more than happy to follow up on this request. I have actually come into contact with an ontologist(home brewer) and another information scientist (beer geek). We've put together a plan to create an abstract pitch for a presentation at a the semantic technology conference in NYC in October (assuming the conference happens... the last one was relatively sparse).

So, I'm more than happy to start posting on this thread links to the ontologies themselves (or questions about how they should be), links to the website, and various versions of the paper. It's much more productive if we steer this thread back in that direction.

At the end of the day, I suppose I'm creating something that may be of use to everyone or no one. Or, it may fail altogether. But that's the nature of inquiry. If we took everything at face value, there wouldn't be much discovery in the world

 

[orangehero]

The BJCP descriptions don't use a consistent and shared beer-cabulary? Like shared descriptors (malty, sweet, fruity, spicy, etc) that are variously qualified for each style?

As I understand it, it's not an attempt to quantify vague and/or subjective concepts, but only to establish relationships between them in the context of beer styles.

Sorry if I missed it, but what is the use again?

 

[MalFet]

Except they don't... nor does any other beer style guide. Style guides contain the descriptions themselves, not a shared vocabulary for making such descriptions. Aside from that, I have no great love for classification. That's a taxonomist's job, and I very much appreciate the work those folks do... but it ain't me. Ontological modeling is not about classification. It is about defining exactly what is means to be a member of a class, and what your relationships are to things of other classes.

I understand what ontological modeling is for, but I suspect you might not appreciate the nature of the senses. Just to be clear, your goal is to create a formalization of perceptual phenomena?

Have you read any of the literature on perception? There are some very basic and intractable facts about how perception works that make what you are proposing here impossible from the outset. Stan Hieronymus's new book on hops actually has a whole chapter on the issue. I'd recommend checking it out if you can. It's an interesting read.

 

[StMarcos]

I'm not an information scientist but I think I have a grasp of the concept in relation to beer, and what the op is attempting to accomplish.

With the BJCP thing.... they have trademarked descriptors so the op is simply asking if they can use them as they complete their own studies.

I do see that the op has not used any obvious examples to help elucidate their project. I understand the academic nature of the project, and that it could be a fun/interesting/thought-provoking. I'm all for it. It would help the op's stance if they could state some hypothetical examples of how the project could progress if continued for some months/years. I think that some readers will want to jump too quickly to 'useful' outputs of such a project, but I see no such need for those. Maybe we could learn something about the concept of 'maltiness.' In what respects is it attributed to fuller base malts (M.O.), larger % of melanoiding rich malts (munich), higher o.g. (barleywines), lovibond of crystal/caramel use, or viscosity increasing adjuncts (rye)?

 

[Yuri_Rage]

So, I'm more than happy to start posting on this thread links to the ontologies themselves (or questions about how they should be), links to the website, and various versions of the paper. It's much more productive if we steer this thread back in that direction.

Do this. You just admitted that you're not even sure exactly what you're trying to quantify. Some examples of your work thus far would help us all to understand your venture.

 

[podz]

Protege is open source and pretty much the de-facto tool for modeling ontologies. It's written in java, so it runs on any machine. You can export your ontologies for import into semantic reasoners such as Apache Stanbol.

I would roughly start with something like this:


thing -> liquid -> beverage -> grain -> fermented -> (branch) top | bottom -> blah

 

[beaksnbeer]

Sounds like taking a simple word like "malty" and using 10+ words to describe what malty is. Like we don't over think enough now we feel the need to complicate and add jargon. Why not just get the lawyers involved and write this up in "legalize" (rant over)

 

[passedpawn]

I would roughly start with something like this:


thing -> liquid -> beverage -> grain -> fermented -> (branch) top | bottom -> blah

Ah, so that's what we're talking about? Looks just like the usual hierarchical classification used in biology. I'd like to see that applied to beer.

 

[podz]

Ah, so that's what we're talking about? Looks just like the usual hierarchical classification used in biology. I'd like to see that applied to beer.

It's not strictly hierarchical; there can be relations between branches.

For example, Alt would be in the ale branch but related to lager because of how it's made. With an ontology loaded into a semantic reasoner, there would thus be a way to mathematically compute that Alt is closer to Weissbier than IPA is. I mean, as a person with a brain we know this via our innate cognitive abilities but computers have no cognitive abilities - they need ontologies and semantic reasoning algorithms in order to derive relationships and form conclusions.

 

[Yuri_Rage]

The more I think I understand the scope of the project, the less interested I become. Is there a need for my computer to "understand" beer? Your example relating Alts, Weissbiers, and IPAs is silly to me. Since we all have the cognitive ability to innately understand that relationship, why would I have to (or want to) quantify and prove it?

 

[Piratwolf]

It's not strictly hierarchical; there can be relations between branches.

For example, Alt would be in the ale branch but related to lager because of how it's made. With an ontology loaded into a semantic reasoner, there would thus be a way to mathematically compute that Alt is closer to Weissbier than IPA is. I mean, as a person with a brain we know this via our innate cognitive abilities but computers have no cognitive abilities - they need ontologies and semantic reasoning algorithms in order to derive relationships and form conclusions.

Understood.

Computers don't drink beer.

Are we really so confused by beer that we need a formal (impossible?) ontology?

I understand Ars Gratia Artis, but I can't see how this improves my enjoyment of or making of beer.

 

[beaksnbeer]

Sounds like something the Gov. would waste millons on in grant money to prove that all beer has water, yeast hops, and barley. A lot of time to relate beers, that share one or more ingredients are similar, who knew.

 

[biochemedic]

It's not strictly hierarchical; there can be relations between branches.

For example, Alt would be in the ale branch but related to lager because of how it's made. With an ontology loaded into a semantic reasoner, there would thus be a way to mathematically compute that Alt is closer to Weissbier than IPA is. I mean, as a person with a brain we know this via our innate cognitive abilities but computers have no cognitive abilities - they need ontologies and semantic reasoning algorithms in order to derive relationships and form conclusions.

Oh, wait, you mean like this:

 

[beaksnbeer]

Oh, wait, you mean like this:

Great there goes the grant money.

 

[podz]

The more I think I understand the scope of the project, the less interested I become. Is there a need for my computer to "understand" beer? Your example relating Alts, Weissbiers, and IPAs is silly to me. Since we all have the cognitive ability to innately understand that relationship, why would I have to (or want to) quantify and prove it?

Yeah, I didn't suggest this project, just explained what they are trying to do and gave an example of the tools to use and how to go about getting started.

That said, don't imagine that something is useless just because it's obvious to you and you can't see a use case. I can easily imagine, for example, medicine evolving to the point where we rely less and less on chemicals and a sick person gets a blood test taken and the calculated prescription is to have a weissbier, pomegranate juice, and almonds each day for a month in order to correct such and such dietary deficiencies.

Having ontologies in place just lays the building blocks for further research and gives the ability to interpolate and understand different types of things. Just because you understand beer doesn't mean that you understand fruits, nuts, or nutrition (maybe you do). Somebody might have a need to understand all four of those things while doing geographical and ethnic interpolation in order to figure out why people are sick even though they follow government eating guidelines.

Strictly regarding creating a beer ontology: somebody will do it anyway. And they might even patent it and put heavy restrictions on it's usage. Unless some group does it first and makes it open source. That might be the best reason to be interested in this project.

 

[MalFet]

Strictly regarding creating a beer ontology: somebody will do it anyway. And they might even patent it and put heavy restrictions on it's usage. Unless some group does it first and makes it open source. That might be the best reason to be interested in this project.

In all earnestness, why I should be concerned about somebody patenting a beer ontology? It's not as if there's a right answer here. It'd just be a list of descriptors that some dude decided were important.

Heck, the BeerSmith library format is a (presumably) patented beer ontology. It's not really standing in the way of my doing anything with beer.

 

[motobrewer]

beerxml.

end of project.


ontology!!!1

 

[TwoRiders]

it is impossible. As has been said before. Quantifying a perception is impossible for any but the individual.

 

[jrubins]

it is impossible. As has been said before. Quantifying a perception is impossible for any but the individual.

OH DEAR GOD, PLEASE DON"T TELL THE BJCP, OR THEY"LL BE OUT OF WORK!!!1111

I keed, I keed,
But in seriousness, it's not impossible (or even highly improbable). With a specific enough framework, it can be done.

Here's why you'd want to do it - A) because you can B) because it's there and C) because it would help provide a framework for the building and judging of beers in a stricter manner, preventing some 'sloppiness' in the formulation of judging and recipe building.

How you do it is by relating certain characteristics to certain styles - e.g. wit- type of: beer made of: wheat malt, barley malt, blah blah blah
has characteristics of: blah blah blah
has components: xyz esters, foo malt profile, bar alcohol content

I'm probably not doing the best job explaining it, but if you approach the problem rigorously, you can form a pretty good ontology. Once you have the framework, you can do fun stuff like: "I want xyz flavor profile". And get an answer like "use foo malt bill, with bar hop schedule, and baz yeast type".

Anyhoo, food (or beer) for thought...

 

[MalFet]

I'm probably not doing the best job explaining it, but if you approach the problem rigorously, you can form a pretty good ontology. Once you have the framework, you can do fun stuff like: "I want xyz flavor profile". And get an answer like "use foo malt bill, with bar hop schedule, and baz yeast type".

Anyhoo, food (or beer) for thought...

Here's a much more modest ambition than the one described here: start a thread and see if you can get everyone to agree on a set of terms that describe what Simcoe tastes like -- objectively, unambiguously, and exhaustively.

If you can't do it (and, to be blunt, I'll bet you can't), the sheer insanity of trying to formalize the perception of beer in its entirety should be obvious.

 

[AmandaK]

Here's a much more modest ambition than the one described here: start a thread and see if you can get everyone to agree on a set of terms that describe what Simcoe tastes like -- objectively, unambiguously, and exhaustively.

If you can't do it (and, to be blunt, I'll bet you can't), the sheer insanity of trying to formalize the perception of beer in its entirety should be obvious.

I think Simcoe smells like a bunch of cats got high and then pissed on a pine tree. Can we put that in the ontology? :cross:

 

[passedpawn]

I think Simcoe smells like a bunch of cats got high and then pissed on a pine tree. Can we put that in the ontology? :cross:

Maybe Malfet is wrong, because I agree 100% with that.

 

[Coff]

I have no clue what any of you people are talking about.

 

[blakelyc]

Is there a need for my computer to "understand" beer? Your example relating Alts, Weissbiers, and IPAs is silly to me. Since we all have the cognitive ability to innately understand that relationship, why would I have to (or want to) quantify and prove it?

Because it is research, and this is how we learn to make new and interesting things. Just because something doesn't seem interesting to you doesn't mean it has no utility.

Oh, wait, you mean like this:

Nope, that's just a confusing representation of a taxonomy.

beerxml.

end of project.

Very no... BeerXML is a data interchange standard, but thanks to this comment I have added BeerXML mapping to the project. I'm not a huge fan of how they represented units of measure in the standard, and they do rely a lot on text in tag contents rather than attributes, but those things can be accommodated. A proposed beer ontology could respond to queries in a beerXML way where that makes sense.

I think Simcoe smells like a bunch of cats got high and then pissed on a pine tree. Can we put that in the ontology?

No facts are invalid

Protege is open source and pretty much the de-facto tool for modeling ontologies. It's written in java, so it runs on any machine. You can export your ontologies for import into semantic reasoners such as Apache Stanbol.

I would roughly start with something like this:


thing -> liquid -> beverage -> grain -> fermented -> (branch) top | bottom -> blah

I'm very much familiar with protege, but my personal tool of choice is TopBraid Composer. I know a lot of OWL purists dismiss it as being to RDF-y, but I chalk that up to the sectarianism that still runs rampant through the semantic tech community.

For tools, here's what I'm thinking:
1.) TopBraid composer (free edition, not open source) to model and output in TTL formatted OWL2.
2.) Pellet 2.8 reasoner (open source)
3.) Stardog RDF store (closed-source, but I think I can score an open source project license for it).
4.) Surface everything to the public using Semantic Media Wiki.

Using an RDF store at all is debatable, but Stardog has Pellet embedded so you get a twofer.

As for the ontology itself, I'll probably propose factoring it into chunks. Since we are talking about one knowledge domain (for now), I also think we can short circuit a lot of stuff so long as we link interesting LOD sources (dbpedia, geonames, etc).

Thoughts?
-b

 

[TwoRiders]

OH DEAR GOD, PLEASE DON"T TELL THE BJCP, OR THEY"LL BE OUT OF WORK!!!1111

I keed, I keed,
But in seriousness, it's not impossible (or even highly improbable). With a specific enough framework, it can be done.

Here's why you'd want to do it - A) because you can B) because it's there and C) because it would help provide a framework for the building and judging of beers in a stricter manner, preventing some 'sloppiness' in the formulation of judging and recipe building.

How you do it is by relating certain characteristics to certain styles - e.g. wit- type of: beer made of: wheat malt, barley malt, blah blah blah
has characteristics of: blah blah blah
has components: xyz esters, foo malt profile, bar alcohol content

I'm probably not doing the best job explaining it, but if you approach the problem rigorously, you can form a pretty good ontology. Once you have the framework, you can do fun stuff like: "I want xyz flavor profile". And get an answer like "use foo malt bill, with bar hop schedule, and baz yeast type".

Anyhoo, food (or beer) for thought...

Beer is to unrestrained and to fluid(pardon the pun) to do something like this with. Wine which is being used as the example has been elitized and is only recently begun to come down to the masses (past 15-20 years or so). SO it has been tightly controlled and quantified using judgement scales that were predetermined by a governing body.

With Beer styles being built and changed on a weekly basis it is almost impossible to do this project unless everything is frozen and the project has a base to start with. Besides the fact the interpretations of beers are dependent on geographic location and water type.

Can it be done? Yes
But it's going to take input from EVERYONE involved in beer creation and judgement as well as some consumers.

SO the scope of the project is MASSIVE, so maybe saying Impossible is harsh but it is a scope of work so large I would just open another homebrew and leave it be.

 

[motobrewer]

I'm waiting for the black ipa class to be created.

Or, is it the cascadian black ale?

 

[AmandaK]

I'm waiting for the black ipa class to be created.

Or, is it the cascadian black ale?

I want the 'American Sour' category.

 

[Milan37]

Ok Wittgenstein (just kidding).

How about you post a very small example of what you're describing? Doesn't have to be the whole finished work, but a small subset would help everyone visualize what you're working on.

Or, even an example from another field (accessible to those of us not named Russell or Frege) would help. I sense that a lot of us have no way to picture what you're trying to accomplish.

 

[Yuri_Rage]

Because it is research, and this is how we learn to make new and interesting things. Just because something doesn't seem interesting to you doesn't mean it has no utility.

Example(s)?

When I first read about your concept, I got the idea that an ontology was a human readable quantification of vague concepts. That would be VERY interesting. Now I think I understand that an ontology is a complicated word for what appears to be a database-type structure. "Reading" it would be an exercise in tedium.

If the utility is to make software capable of "understanding" ingredients and styles, that software would be pretty useless. A computer could certainly be made to create a style-perfect recipe, but homebrew recipes have few enough ingredients that a basic understanding of the process should give the brewer a very good idea of how that recipe will turn out. Computer validation of the concept is unnecessary.

How about you post a very small example of what you're describing?

This, too.