Simplify AOF Categories +8
Problem: AOFs are too complicated (379 categories rolled into 46 top levels) for users, hinder globalization (too much to translate), and are not very scientific (i.e. not exhaustive and mutually exclusive).
Solution: Could we use an existing taxonomy (e.g. Earth Charter) to bring some rationalization to the AOFs? Merge and purge down to say 50 AOFs at the bottom level rolled up into 4 to 12 at the top level?
How I can help: I would be happy to assist in this process (by angusparker)
____________________________________________________________________________________
Orginal Post: A few thoughts about WiserEarth categories aka Areas of Focus :
Why reinvent vocabulary/taxonomy? This is a highly time-consuming task. There are published taxonomies and vocabularies out there The purpose of vocabulary is to find and connect resources inside WiserEarth, but also to be found and eventually linked from outside. I have in mind at some point that WiserEarth content should be made available as Linked Data. Singularly why not use Wikipedia categories, as represented in DBpedia. I guess many, if not most AoF, both existing and required, are defined there. Just a few examples below.
http://dbpedia.org/resource/Category:Agriculture
http://dbpedia.org/resource/Organic_farming
http://dbpedia.org/resource/Permaculture
Benefits to use such URIs to identify categories
- Ability to extend to related categories
- Ability to connect to other linked data on the same subject (outreach on the Semantic Web)
- Access to multilingual descriptions available through those URIs
Cross posted in WiserEarth Suggestions and WiserEarth Area of Focus Suggestions
Comments (1 - 20 of 77)
|
|
I recently received this message from Honore and thought it would be good to share on this discussion: "The AoF Selection system is well known for having poor usability. Cutting down the number of categories likely doesn't go far enough in addressing what it should be doing - which is collecting information about a user's interests. An example of a better model is Stumbleupon - it uses a hybrid between selecting from a drop-down menu and allowing users to freely type in keywords, so long as they're separated by commas. The fixed system is great for connecting with DBpedia for keyword translation, and the text box provides freedom for the user to build up a "folksonomy"." |
|
|
Hi Jaana, thanks for jumping into the discussion. Great to see you here.
|
|
Hi all! I finally had time to look into this discussion and I have to say that I am super happy with the content :D ! Let me introduce myself - I am Jaana from Finland and I work at the Earth Charter International Secretariat as an International Youth Facilitator. In other words, I continued the job Dominic left in February (Dominic has given his thoughts on this discussion ealier this and last year). I wanted to let you know that what ever support you need from us (ECI Secretariat) - let us know! We can try to help you out to make this happen! For me the Option A looks easier to understand - but I am not too familiar with WE yet so I need to browse this site a bit more before I can give you my final opinion :) ! I am so happy for this! Let us know if we can something for this to help you out! Jaana |
|
|
This comment was removed by a WiserEarth editor for the following reason:
Duplicate entry |
|
|
Whoops, I meant @Pierre, not @Peter :)
|
|
|
@Bret: I like the modifications. I think whatever interface we end up using, which needs careful thinking and testing, these modifications should be taken into consideration. Let's think about the options and possibilities some more. They certainly look interesting. Perhaps Pierre can present us with some fresh perspective when he return to office? @Peter: Thanks for joining us here Peter. Looking forward to hear your thoughts. And how nice to finally see a smile (photo) there :) |
|
I will have a more thorough look at this, on return to my office, after June5th. Will that be all right ?
|
|
Just a possibility, but maybe we could have our cake and eat it too. Consider Bowo's four-tiered taxonomy--3 top categories to 12 EC categories to 53 EC subcategories to 379 WiserEarth AoFs--implemented very much like this flyout menu but with the following modifications:
Why I bring this up is that it might actually be a more educational and, thereby, more accurate solution. For example, when I went through Bowo's four-tiered system for a group I am involved in, I learned that our previous categorizations were not that accurate: Organization AoFs: Peace and Peace Building | Media and Communication | Democratic Reform | Distributive and Economic Justice | Social Justice Education Group AoFs: Environmental Justice | Distributive and Economic Justice | Social Justice Education | Human Rights and Civil Liberties | Law and Policy Reform | Human Rights Protection The most appropriate category for our group was the EC subcategory Peace Philosophy and Practice. However, the closest WE 379 AoF would be Peace and Peace Building, which was only listed for the organization page. In fact, the other eight AoFs listed for the organization and group pages are not that appropriate, given the group's actual practice. Other benefits with this possibility:
|
|
|
From our discussion, here's a basic action plan, with our discussed options, for Stage One.
Though I am for Option B, I am most for Option A. It is the simplest to implement/understand/use and would most redress the concern of this article. Further, it is forwards compatible with Option B--in case Option A simplifies things too much, which I doubt, especially with the original 379 WE AoFs in place as tags. I am leaning away from Option X, as it would mix ever-expanding keywords with the Blessed Unrest original AoFs, blocking backwards compatibility. Although having both a tags field and a keywords field seems redundant, perhaps in the future we can merge them or eliminate one of them. But I think this topic would be better discussed under Stage Two. |
|
@ Brett: As you suggest we will probably have to tackle this in a two stage way - AOF/Earth Charter mapping, then DBPedia.
|
|
|
Thanks for the clarifications; I did misunderstand how the DBpedia taxonomy was to be used. However, I still have a couple of concerns.
When the current AoFs are translated to the DBpedia taxonomy, the resulting number of effective AoFs (resulting tags) will likely be reduced. This is probably a good thing but it brings up Bowo's point: "It seems that some level of discontinuity is inevitable with this option..." I wonder if this is a real concern or not. Probably, since the tagging translation will be done behind the scenes, this will not be a real issue though.
That seems to leave my last concern. Will new tags be allowed to be created? If so, doesn't that avail the tagging system to "hundreds of thousands of crazy categories"? Tied into this is the question of whether keywords are going to be converted to the same tagging system. Hence, perhaps we should hold off in dealing with the DBpedia aspects (besides for under-the-hood tag translations to the existing AoFs taxonomy--possibly), to just move forward with reducing the AoFs: to solve the fundamental issue that this article addresses that much faster. |
|
@Bret As Bowo mentions correctly, it has never be proposed to replace WE taxonomy by DBpedia taxonomy with its hundreds of thousands of crazy categories, but two things instead. #1. To map each of the 12-50 WE taxonomy entries, whichever they happen to be at the end of the day, to one or more DBpedia resource, carefully chosen. Can be DBpedia "category" or DBpedia "resource". The important point is to hook the WE taxonomy to entries in the linked data cloud, and DBpedia is rich enough to tap in. #2. Use DBpedia tags instead of free tagging, a la faviki. #1 is built-in, under the hood, and the final user does not know about it. #2 would be made using auto-completion, but I agrees that maybe it's too rich for the task at hand, and is at risk to be abused, misused or not used. I like Faviki, but it has not been a killer app so far. Maybe opencalais is a better choice. It does not put any burden on the user. So we would have human categorization on a simple taxonomy, mapped to other entries in DBpedia or any other resource we see fit in the SemWeb cloud, and automatic semantic tagging of the content with opencalais. Why not. @Camilla Yes the format for the categories-to-DBpedia mapping you propose is OK @Bowo Searching DBpedia maybe difficult, but you can search Wikipedia instead, and as soon as you have found a relevant article, you have just to change the namespace to get the DBpedia resource, e.g., http://en.wikipedia.org/wiki/Permaculture > http://dbpedia.org/resource/Permaculture |
|
|
Looks like thinking is firming up ... ;-)
|
|
|
@Bret: The linkup to DBpedia does not affect users directly. What they see will be a selection of AoFs (12 to start with, and 50 or so if they want be more specific) and manual tag entry. In all WE pages, what's visible will also be just that (several AoFs and several tags). The DBpedia linkup serve to connect things under the hood. Think google indexing (extremely complex) and google interface (simple). What's relevant to users is the interface. If we can automate the linkup to semantic web (e.g. with something like OpenCalais), all the better since all the complexity will be dealt with automagically.
On the two-tiered selection. Actually, that could work quite well. People would either tag using 3 of the 12 top level AoFs, or choose from 3*(50/12)=roughly 12 second level AoFs. Still looks simple enough. But we'll have to test under real world usage conditions. Earth Charter use would be more useful and meaningful if we incorporate the second layer (50 or so sub-principles) as well. Remember, that 12 steps was a sketch open to change, not something set in stone. |
|
|
http://opencalais.com/ should we consider using this at all for the semantic web aspect as they have an automatic approach?
|
|
|
I believe the DBpedia idea is not viable; it has over 400,000 categories (as far as I can tell). This would introduce psychotic complexity, where WE already has too many AOF categories: that is the thing we are trying to simplify! Just linking/mapping to DBpedia categories might be a good idea. However, that seems independent to the key issue we are trying to solve.
I really think we need to keep the AOFs simple. A multi-tiered system is a complex concept for those not that computer savvy: for probably the majority of the WE community. For these reasons, I again promote Wo's idea of just having 12 categories. That's manageable and simple for all users.
Wo's 12-step plan might be scarily long, but the first three steps seem to be spot on. Although this issue is a big deal, it has been extremely well-debated and around for too long not for us to consolidate our opinions and take action. The WE community is suffering right now because of this significant problem. Can we agree to Wo's first three steps and begin to take action on them? The rest of the steps we can continue to discuss.
Forward ho! |
|
|
Great to see this moving forward a little bit more.
On option 1: It maybe a good idea to enable a two-tiered Earch-Charter based taxonomy as suggested (12 top level, 50 or so second level), where each layer is selectable (as Bret suggested), but not both 'parent' and 'child' item. Should provide (initial) granularity while allowing for simplicity. Full granularity can be achieved with tags/keywords. I would support each AoF retaining a portal page and also give 'portal' pages for tags/keywords, so we can link all to DBpedia. It seems that some level of discontinuity is inevitable with this option, and it will take more work to implement than other options.
On option 2 and 3: Adds complexity and thus, not improve usability. So, doesn't look like a solution for the problem at hand.
On option 4: Could be an interim solution, but I wouldn't advocate it for the long run since it reduces flexibility. I think we want simplicity, flexibility and granularity at once. Yes, not easy!
Having said that, I agree that we should proceed with great care as this will effect the entire site. And whatever we do, the main goal (I think) is to make it easier for people to add content to WiserEarth. Tagging, finding and recommending content comes later / at a lower priority.
@Bernard: Thanks for reminding us of Faviki again. Yes, their tag correction, suggestion and translation feature looks like a real solution to me too! Some questions:
|
|
|
OK - I'll get onto this.
|
|
|
@Camilla Regarding the mapping of existing AoFs to DBpedia, what we (even you) can do is :
Examples of rows in this table aof/113:Permaculture dbpedia:Permaculture aof/345:Rural Farming Communities dbpedia:Rural_community_development, dbpedia:Agricultural_cooperative The second example (even if questionable) to show that one AoF may be linked to several DBpedia entries. When this spreadsheet is completed, I will transform it into RDF using the SKOS matching vocabulary, and we can dwelve into the gory details of RDF publication best practice.
|
Welcome Jaana!
Very glad to have your energy and support in this!
Looking forward to hear more and happy to help you feel at home here :)
@Camilla:
On Honore's comment: If we give each keyword a URL/page (displaying search results for that keyword for example), then from what Bernard said so far, it should also be possible to link them up with the semantic web.