OUseful.Info, the blog…

Whilst on holiday a couple of weeks ago, we took an audio tour on an open top bus. I missed a significant chunk of the tour because my headphones broke after the first couple of minutes (the cable was only a tiny, easily snapped strand of copper thick… I guess the cost of copper played a part in that…?!) but it got me thinking (again) about geofenced audio tours.

If the term is new to you, geofencing refers to a technique where you put a notional boundary round a location and capture the GPS coordinates of the boundary; if the boundary is a regular shape, you can calculate the boundary co-ordinates via a simple formula. For example, if you have a circular region 1 km wide about a point, the geofence is defined by the circumference (or as Terry Pratchett would say,m the circumfence) of the circle, centred on the point of interest and with a radius of 0.5km. The geofenced region then lies within this circle.

For audio tours, what this means is that you can go for a walk, or a ride, and as an when you within the confines of the geofence around an object, a commentary about that object can play out for you. I had a quick look around for apps that might support the creation of such a tour, but there didn’t seem to be much out there. Here’s what I did find though:

- Hackney Hear – An interactive GPS-triggered audio tour of Hackney; creative arts project to develop an audio tour app around the London Borough of Hackney; due out in January 2012. Looks exciting (will they open the code, I wonder?)
- NoTours – Augmented Aurality for Android: this site includes an editor for creating your own geofenced audio tours, and a demo app that lets you play a tour with up to 10(?) audio locations marked. The site espouses open source principles but I couldn’t see a link to the code anywhere? (The site name actually reminded me of the idea of a misguided walk;-)
- Geovative geotours: a commercial offering, though the free plan suggests you can create up to three tours.

Unfortunately, I’ve not had a chance to play with any of them yet…

(If you know of any others, particularly open source apps and tour creators, please let me know via the comments.)

One thing I’d quite like to do one day is create an app that I can listen to on the motorway that will play out stories about the places I’m passing. (I wonder if I could build such a thing in Android App inventor?). As a first pass, I could imagine querying something like Wikipedia (e.g. using this new wikilocation api) to pull back articles relating to points of interest near my current location, and then using text to speech to read a selected article out.

(Note to self: I think Audioboo allows geotagged audio uploads. Does it also support geo-based searches? This could provide a secondary source of commentaries…?)

It might also be fun to try to create tours for bus routes, e.g. as identified from the wonderful new MySociety site, FixMyTransport (though adding such functionality to that site directly would be the sort of feature creep that I think MySociety sites always try to avoid: How to create sustainable open data projects with purpose).

And finally, whilst on the question of geofences, and what actually brought them back to mind today: flickr has just opened up geo-privacy fences: Introducing geofences on Flickr!:

Geofences are special locations that deserve their own geo privacy settings. For example, you might want to create a geofence around the your “home” or “school” that only allows “Friends and Family” to see the location of the photos you geotag in that area. So the next time you upload a photo with a geotag in the radius of a geofence, it will follow the default geo privacy you’ve designated for that hotspot.

Clever ;-)

PS here’s an alternative take on “geofences” (h/t @AidanBaker for the reminder about this art piece:-)

The artists involved created a rod that showed the wifi strength at a particular location, then, using time lapsed photography, took it for a walk: Immaterials: Light painting WiFi

PPS fwiw, you can run spatial queries over geo-data hosted in Google Fusion Tables: Search your geo data using spatial queries from Fusion Tables Spatial queries can be run “via the WHERE clause. The syntax is: ST_INTERSECTS(, ) For , use a in your table that contains location data. For , use one of the following: ” I haven’t tried it yet, but this approach looks amenable to geofence style query activity within regularly bounded regions.
A weaker spatial query form (that cannot be combined with ST_INTERSECTS conditions) to use ORDER BY based on distance. “The syntax is: ST_DISTANCE(, ) For , use a in your table that contains location data. Listing the as a is optional when using ORDER BY ST_DISTANCE. ORDER BY ST_DISTANCE cannot be combined with any of the ST_INTERSECTS conditions.”

PPPS and finally, just in case, here’s a link to the code repository for the Google MyTracks Android app; it may contain useful code snippets for any homebrew native Android app…

I spent a chunk of today trying to get my thoughts in order for a keynote presentation at next week’s The Difference that Makes a Difference conference. The theme of my talk will be on how visualisations can be used to discover structure and pattern in data, and as in many or my other recent talks I found the idea of Anscombe’s quartet once again providing a quick way in to the idea that sometimes the visual dimension can reveal a story that simple numerical analysis appears to deny.

For those of you who haven’t come across Anscombe’s quartet yet, it’s a set of four simple 2 dimensional data sets (each 11 rows long) that have similar statistical properties, but different stories to tell…

Quite by chance, I also happened upon a short exercise based on using R to calculate some statistical properties of the quartet (More useless statistics), so I thought I’d try and flail around in my unprincipled hack-it-and-see approach to learning R to see if I could do something similar with rather simpler primitives than described in that blog post.

(If you’re new to R and want to play along, I recommend RStudio…)

Here’s the original data set – you can see it in R simply by typing anscombe:

   x1 x2 x3 x4    y1   y2    y3    y4
1  10 10 10  8  8.04 9.14  7.46  6.58
2   8  8  8  8  6.95 8.14  6.77  5.76
3  13 13 13  8  7.58 8.74 12.74  7.71
4   9  9  9  8  8.81 8.77  7.11  8.84
5  11 11 11  8  8.33 9.26  7.81  8.47
6  14 14 14  8  9.96 8.10  8.84  7.04
7   6  6  6  8  7.24 6.13  6.08  5.25
8   4  4  4 19  4.26 3.10  5.39 12.50
9  12 12 12  8 10.84 9.13  8.15  5.56
10  7  7  7  8  4.82 7.26  6.42  7.91
11  5  5  5  8  5.68 4.74  5.73  6.89

We can construct a simple data frame containing just the values of x1 and y1 with a construction of the form: data.frame(x=c(anscombe$x1),y=c(anscombe$y1)) (where we identify the columns explicitly by column name) or alternatively data.frame(x=c(anscombe[1]),y=c(anscombe[5])) (where we refer to them by column index number).

x y
1 10 8.04
2 8 6.95
3 13 7.58
4 9 8.81
5 11 8.33
6 14 9.96
7 6 7.24
8 4 4.26
9 12 10.84
10 7 4.82
11 5 5.68

A tidier way of writing this is as follows:
with(anscombe,data.frame(x1Val=c(x1),y1Val=c(y1)))

In order to call on, or refer to, the data frame, we assign it to a variable: g1data=with(anscombe,data.frame(xVal=c(x1),yVal=c(y1)))

We can then inspect the mean and sd values: mean(g1data$xVal), or sd(g1data$yVal)

> mean(g1data$xVal)
[1] 9
> sd(g1data$xVal)
[1] 3.316625
>

To plot the data, we can simply issue a plot command: plot(g1data)

It would be possible to create similar datasets for each of the separate groups of data, but R has all sorts of tricks for working with data (apparently…?!;-) There are probably much better ways of getting hold of the statistics in a more direct way, but here’s the approach I took. Firstly, we need to reshape the data a little. I cribbed the “useless stats” approach for most of this. The aim is to produce a data set with 44 rows, and 3 columns: x, y and a column that identifies the group of results (let’s call them myX, myY and myGroup for clarity). The myGroup values will range from 1 to 4, identifying each of the four datasets in turn (so the first 11 rows will be for x1, y1 and will have myGroup value 1; then the values for x2, y2 and myGroup equal to 2, and so on). That is, we want a dataset that starts:

1 10 9.14 1
2 8 8.14 1

and ends:

43 8 7.91 4
44 8 6.89 4

To begin with, we need some helper routines:

- how many rows are there in the data set? nrow(anscombe)
- how do we create a set of values to label the rows by group number (i.e. how do we generate a set of 11 1′s, then 11 2′s, 11 3′s and 11 4′s)? Here’s how: gl(4, nrow(anscombe)) [typing ?gl in R should bring up the appropriate help page;-) What we do is construct a list of 4 values, with each value repeating nrow(anscombe) times]
- to add in a myGroup column to a dataframe containing x1 and y1 columns, set with just values 1, we simply insert an additional column definition: data.frame(xVal=c(anscombe$x1), yVal=c(anscombe$y1), mygroup=gl(1,nrow(anscombe)))
- to generate a data frame containing three columns and the data for group 1, followed by group 2, we would use a construction of the form: data.frame(xVal=c(anscombe$x1,anscombe$x2), yVal=c(anscombe$y1,anscombe$y2), mygroup=gl(2,nrow(anscombe))). That is, populate the xVal column with rows from x1 in the anscombe dataset, then the rows from column x2; populate yVal with values from y1 then y2; and populate myGroup with 11 1′s followed by 11 2′s.
- a more compact way of constructing the data frame is to specify that we want to concatenate (c()) values from several columns from the same dataset: with(anscombe,data.frame(xVal=c(x1,x2,x3,x4), yVal=c(y1,y2,y3,y4), mygroup=gl(4,nrow(anscombe))))
- to be able to reference this dataset, we need to assign it to a variable: mydata=with(anscombe,data.frame(xVal=c(x1,x2,x3,x4), yVal=c(y1,y2,y3,y4), mygroup=gl(4,nrow(anscombe))))

This final command will give us a data frame with the 3 columns, as required, containing values from group 1, then group 2, then groups 3 and 4, all labelled appropriately.

To find the means for each column by group, we can use the aggregate command: aggregate(.~mygroup,data=mydata,mean)

(I think you read that as follows:”aggregate the current data (.) by each of the groups in (~) mygroup using the mydata dataset, reporting on the groupwise application of the function: mean)

To find the SD values: aggregate(.~mygroup,data=mydata,sd)

Cribbing an approach I discovered from the hosted version of the ggplot R graphics library, here’s a way of plotting the data for each of the four groups from within the single aggregate dataset. (If you are new to R, you will need to download and install the ggplot2 package; in RStudio, from the Packages menu, select Install Packages and enter ggplot2 to download and install the package. To load the package into your current R session, tick the box next to the installed package name or enter the command library("ggplot2").)

The single command to plot xy scatterplots for each of the four groups in the combined 3 column dataset is as follows:

ggplot(mydata,aes(x=xVal,y=yVal,group=mygroup))+geom_point()+facet_wrap(~mygroup)

And here’s the result (remember, the statistical properties were the same…)

To recap the R commands:

mydata=with(anscombe,data.frame(xVal=c(x1,x2,x3,x4), yVal=c(y1,y2,y3,y4), group=gl(4,nrow(anscombe))))
aggregate(.~mygroup,data=mydata,mean)
aggregate(.~mygroup,data=mydata,sd)
library(ggplot2)
ggplot(mydata,aes(x=xVal, y=yVal)) + geom_point() + facet_wrap(~mygroup)

PS this looks exciting from an educational and opendata perspective, though I haven’t had a chance to play with it: OpenCPU: a server where you can upload and run R functions. (The other hosted R solutions I was aware of – R-Node – doesn’t seem to be working any more? online R-Server [broken?]. For completeness, here’s the link to the hosted ggplot IDE referred to in the post. And finally – if you need to crucnh big files, CloudNumbers may be appropriate (disclaimer: I haven’t tried it))

PPS And here’s something else for the data junkies – an easy way of getting data into R from Datamarket.com: How to access 100M time series in R in under 60 seconds.

[Note that I might treat this post a bit like a wiki page... Note to self: sort out a personal wiki]

Call is JISC OER3 – here’s the starter for ten (comments appreciated, both positive and negative; letters of support/expressions of interest welcome; comments relating to possible content/themes, declarations of interest in taking the course, etc etc also welcome, though I will be soliciting these more specifically at some point)

Rapid Resource Discovery and Development via Open Production Pair Teaching (ReDOPT) seeks to draft a set of openly licensed resources for potential (re)use in courses in two different institutions through the real-time production and delivery of an open online short-course in the area of data handling and visualisation. This approach subverts the more traditional technique of developing materials for a course and then retrospectively making them open, by creating the materials in public and in an openly licensed way, in a way that makes them immediately available for “study” as well as open web discovery, and then bringing them back into the closed setting for (re)use. The course will be promoted to the data journalism and open data communities as a free “MOOC” (Massive Online Open Course)/P2PU style course, with a view to establishing an immediate direct use by a practitioner community. The project will proceed as follows: over a 10-12 week period, the core project team will use a variant of the Pair Teaching approach to develop and publish an informal open, online course hosted on an .ac.uk domain via a set of narrative linked resources (each one about the length of a blog post and representing 10 minutes to 1 hour of learner activity) mapping out the project team’s own learning journey through the topic area. The course scope will be guided by a skeletal curriculum determined in advance from a review of current literature, informal interviews/questionnaires and perceived skills and knowledge gaps in the area. The created resources will contain openly licensed custom written/bespoke material, embedded third party content (audio, video, graphical, data), and selected links to relevant third party material. A public custom search engine in the topic area will also be curated during the course. Additional resources created by course participants (some of whom may themselves be part of the project team), will be integrated into the core course and added to the custom search engine by the project team. Part-time, hourly paid staff will be funded to contribute additional resources into the evolving course. Because of timescales involved, this proposal is limited to the production of the draft materials, and does not extend as far as the reuse/first formal use case. Success metrics will therefore be limited to volume and reach of resources produced, community engagement with the live production of the materials, and the extent to which project team members intend to directly reuse the materials produced as a result.

Some bits and pieces of the jigsaw that may or may not relate to a bid I’m trying to write… it’s a core dump of base/basic principles and assumptions that I may or may not believe, to see whether or not any of them actually make sense…

- from a tweet a few weeks ago: OERs as “openly licensed resources that educators can reuse in teaching, or that learners can independently discover and learn from”

- an untested suspicion that OERs (whatever they are) are often produced under the tacit assumption that if they are to be reused, they will be selected by an educator and reused in a formal educational setting, or will be discovered (in some magical way) within an OER repository and reused by a student in a formal educational setting.

- a utilitarian view that open educational resources are just resources. That if you can find them, are willing make use of them, and can actually make use of them at least in part how you want to (which may include repurposing them, editing them, etc etc), they are to all intents “open” to you; and that they are educational if they help someone learn something or how to do something (e.g. so that if someone comes to a resource not knowing something, or not knowing how to do something, and after engaging with the resource they do know the something, or how to do it, without having to refer back to the resource, then they’ve learned something about the something…)

- a belief that a resource that can be open and educational (as described above) might also be trivial: a flashcard containing a picture of a dog and the word dog next to it, for example.

- a belief that an OER is only useable or reusable if it fits into someone else’s educational process (someone wanting to know how to do a thing, finds a resource, and ends up learning something about the thing; or a teacher trying to help someone else learn something.)

- an observation that academics often make a distinction between a “proper” academic education, and training (so you know how to design a car from first principles, but have no idea how to actually bend pieces of metal to make one, or even how to drive one;-) (Hmmm: try this: academic education helps you to make distinctions, ask questions, and find alternative ways of doing the same thing; it provides you with (cognitive) tools for abstracting out elements of a problem and solving that problem. Training maybe just shows you a single way of addressing a particular problem in a practical way?)

- a recollection of thoughts from when I last spent some time thinking about the “uncourse” model I used to draft materials for the OU Digital Worlds course: that courses typically have a single linear narrative that is determined prior to the start of the course; that the uncourse had a linear narrative only in the sense that it documented a set of topic explorations by means of a series of blog posts that were posted in serial/linear time as and when I wrote them that documented my (learning) journey through the topic; but that by the use of link structure, multiple other (linear) narratives could be woven through the network of resources that my linear sequence of posts left behind, or linked to themselves; that “self-containment” is often viewed as plus in the production of OERs, so they operate as standalone resources; but that by embedding resources in a network of other resources you potentially make them discoverable through those connected resources, and also provide context derived from those connected resources and the way they are connected; that one model for uncourse blog posts was a template that included: a question, a set of linked resources, and an example answer to the question or reflection on the linked to resources. That the resource might contain both “academic” and “training” components, that could help someone solve a practical problem (the training element) yet also referred to abstractions and reflections contextually relevant to the training element (the academic component). That as such, the resource: a) was appropriate for someone discovering the resource via a “how do I?” search query (the training bit); b) that the resource contained a contextualised example (the training bit) that could be reused by another educator or picked up on by a user who originally arrived looking for an answer to a how-do-I question; c) that the resource contained a contextualisation container (the academic bit) that could be reused by another educator with their own training component; that the uncourse produced resources that were primarily part of the web and could freely be discovered on a post-by-post basis via web search engines by users searching for appropriate terms. That the uncourse resources were not made discoverable (hah!) by virtue of being placed in siloed OER repositories, but were gifted into the web via a public website (a hosted wordpress.com blog) that made no claims about having academic authority or provenance.

Maybe related: @ambrouk’s Connecting people through open content and @kavubob’s Public funding, OER, and Academics – a brief reflection and An OER manifesto in twenty minutes

PS This post was written whilst trying to write a JISC OER3 bid, struggling to find the words and phrasing required to clarify the jumble of thoughts relating to what I (think I’m) bidding, for, and already suspecting that I’ve left it too late to get the signatures/sign-offs and accounting stuff done in time (needless to say, folk are on holiday, and OU deadline is a week before the JISC submission deadline)… Plus the fact the bid still isn’t written and the bits that are pretty weak:-( Sigh… I guess that makes this post a displacement activity, albeit one I thought might help with the bid (it hasn’t:-(. Needs must though – as we’re moving from being publicly funded to being funded by paying customers, I suspect the “public service because I’m publicly funded” argument I use to stave off the guilt and try to justify my activities (to myself at least) won’t work past October 2012, if even until then:-( Marketisation of HE is really going to mess things up, methinks:-(

There’s been quite a flurry of OU job ads out recently in the IT and analytics areas… Here’s a quick round-up:

• Head of IT Development (Curriculum Planning, Finance and HR) (£46,696 – £55,758): “We are looking for a Head of IT Development to manage the Curriculum Planning, Finance and HR team within IT Development.
  “You will manage a team of 15-20 Analysts and Developers, and will have responsibility for the utilisation, skills and development of staff within the team. They must be able to demonstrate leadership and management skills and the ability to drive change to completion. The Head of IT Development also has responsibility for the relationship with business partners using our Curriculum Planning, Finance and HR systems, liaising with stakeholders on a regular basis and takes overall responsibility for the delivery of the IT Development project portfolio for their business area.”
• Head of IT Testing, Information Technology (£46,696 – £55,758): “We are looking for a Head of IT Testing to establish and lead a new IT Testing function within IT Development.
  “The post holder will assess current testing processes, tools and environments then produce and implement an overall testing strategy. The Head of IT Testing will be accountable for the quality of testing for all projects delivered by IT Development. They must be able to demonstrate leadership skills and the ability to drive process and culture change into a busy and challenging environment. The IT Development team deliver a diverse range of projects including multi-million pound systems transformation programmes, large package upgrades and enhancements and key strategic business change projects.
  “Applicants should preferably hold recognised IT Testing qualifications and will be fully conversant with best practice project management and IT development methodologies. They must have demonstrable experience of managing IT testing on large software programmes and of successfully creating and managing IT testing functions.”
• 3x Business Systems Analysts, Information Technology (up to £39,000): “We are looking for experienced systems analysts to work in the Development section of the University’s IT Department, to support and enhance existing business applications and work on business change projects.
  “The University has recently approved a multimillion pound 3 year systems redevelopment programme. IT Development is tasked to deliver this major change programme while continuing to support the existing estate primarily its in-house systems.
  “We need to strengthen our existing 30 strong analyst team with additional experienced analysts. While this will involve a significant amount of in-house development, we expect to increase the use of off-the-shelf software, so experience of package configuration and support would also be useful.”
• IT Project Manager, Information Technology (£36,862 – £44,016): “We are looking for one IT Project Managers to work within a new IT Project Management function within IT Development, reporting to the Head of IT Project Management.
  “You will be responsible for the successful delivery of all projects assigned to them and must be able to demonstrate leadership skills and the ability to drive change to completion.”
• 2x Senior Business Improvement Analysts, CIO’s Office, Information Technology (£36,862 – £44,016): “The Higher Education sector is undergoing substantial change; with restructuring of funding provision and increased competition driving the need for an increasingly customer-centric, agile and efficient service. As a result, we are expanding the Business Process Improvement team whose role is to work with business partners to identify opportunities to improve efficiency and transform processes.
  “We are looking for experienced BPI Analysts with a background in process change and business transformation. Taking problem / ambition statements from our business partners across the organisation, you will work with them to develop future state processes and the roadmaps to implement those changes.
  “Knowledge of process analysis techniques and the Lean methodology is essential. You should also be a confident, experienced facilitator and communicator (used to dealing with senior and board level executives) with strong influencing & negotiation skills and enjoy working in a collaborative environment.”
• Systems Analyst, Information Technology (up to £39,000): “We are looking for experienced systems analysts to work in the Development section of the University’s IT Department, to support and enhance existing business applications and work on business change projects.
  “The University has recently approved a multimillion pound 3 year systems redevelopment programme. IT Development is tasked to deliver this major change programme while continuing to support the existing estate primarily its in-house systems. We need to strengthen our existing 30 strong analyst team with additional experienced analysts. While this will involve a significant amount of in-house development, we expect to increase the use of off-the-shelf software, so experience of package configuration and support would also be useful. A good knowledge of structured development methods, including process and data analysis, is also a prerequisite. You should also enjoy working collaboratively, have a flexible approach and a keenness to develop your professional skills.”
• Information Analyst (£29,972 – £35,788): “We are seeking someone to join our small team of information analysts, providing data and analysis to inform strategic and operational planning, decision making and external reporting requirements.
  “A key part of the role is the preparation and provision of analysis to support student recruitment forecasting and target-setting, as well as providing wider management information and reports in support of institutional and unit planning.
  “If you are interested in joining us, you should have a first degree (or equivalent) in a subject with a strong mathematics and/or statistics component and have experience of data manipulation and reporting using SAS, SPSS or a similar analytical tool.”
• Learning and Teaching Librarian, Library Services (part-time, £29,972 – £35,788 (pro rata)): “We are seeking a qualified Librarian to join our professional team. You will provide specialist support to module and programme teams in the implementation of the Learning and Teaching Strategy, as well as supporting research activity. You will provide information and support to students and tutors, assist in the development and delivery of information skills training and assist in collection management.
  “You will be confident with new technologies and have the ability to excite and engage users in developing information skills and resource-based learning activities to enhance the Open University curriculum. You will understand the pedagogic use of resources and technologies and aware of the issues around digital scholarship.”
• Comms and Relationship Manager (SocialLearn), Knowledge Media Institute (KMi) (£36,862 – £44,016): “The Open University’s Knowledge Media Institute has an opening for a Comms and Relationship Manager on the SocialLearn Project.
  “As the project moves into a new phase where we will be actively building new user communities, the need to develop partnerships will be a key success factor. Also crucial is the need to build an active user base drawn from both internal and external communities. We are, therefore, looking to acquire the services of an experienced professional to develop and manage the various audiences and stakeholders that will either have an interest in SocialLearn or the potential to engage with the system as users or developers.”
• Research Associate (SocialLearn): Social Learning Analytics & Recommender Services, Knowledge Media Institute (KMi): “The Open University’s Knowledge Media Institute has an opening for a Research Associate: Social Learning Analytics & Recommender Services. The SocialLearn project is investigating the future of online social learning, and draws on expertise from across the OU [www.open.ac.uk/sociallearn].
  “Your responsibility will be to use your understanding of analytics and recommendation engines to improve the SocialLearn user experience.”

For more information about the SocialLearn project, see: SocialLearn beta, blog

For how IT fits into the OU structure, see the OU Administrative Substructure Org Chart [PDF]

Courses

If you’re looking for related qualifications/career enhancement pathways, see also:

- CIO Executive Council Pathways
- Course – Network security: “With organisations increasingly dependent on their information systems, there is a greater need to ensure that the underlying network infrastructure being used by these systems is secure. As a leading Cisco Academy, this course includes the Cisco Certified Networking Associate (CCNA) Security course and provides a foundation for preparing for the Certified Information Systems Security Professional (CISSP) qualification.”
- undergrad courses in Computing/ICT (register now to lock in cheaper fees;-)
- undergrad qualifications in Computing and ICT (register now to lock in cheaper fees;-)
- postgrad qualifications in Computing and ICT, postgrad courses in business, postgrad qualifications in Technology and Technology Management

As an OU employee, being able to register on OU undergrad courses for free has been a traditional perk… (“Staff Fee Waivers can be used for any taught course that the University offers: this includes Openings, Arts, Science and Technology Short Courses and Residential School Courses. PGCE/ITT courses are not covered unless you have special approval from your Head of Unit.”) I assume this will continue to be the case when the new fees regime kicks-in…?

Here’s a quick snapshot of a some List Intelligence recipes I’m working on in the area of list and notable individual discovery….

The starting point in each case is a Twitter list in a particular topic area. The list I’m going to use in this example is @jobrodie’s scicomm list, a list of science communicators on Twitter.

The first recipe-in-progress tries to identify other lists in the area… It works as follows:

- for each person on the original list (candidate members), find the other lists they are mentioned on to give a set of candidate lists;
- count the number of candidate members (members of the original list) on each candidate list.

We can now display lists containing N or more people on the original list:

/JoBrodie/scicomm Candidate members: 61 Subscribers: 9
/ebsuckling/science Candidate members: 23 Subscribers: 14
/lisushi/science-world Candidate members: 22 Subscribers: 6
/Frazerbw/science Candidate members: 22 Subscribers: 2
/efficientva/science Candidate members: 22 Subscribers: 7
/cptyler/science Candidate members: 20 Subscribers: 15
/LouWoodley/london-sci-comm Candidate members: 20 Subscribers: 7
/ousefulAPI/solo10 Candidate members: 18 Subscribers: 1
/mocost/london Candidate members: 17 Subscribers: 25
/girlinterruptin/science-science-policy Candidate members: 16 Subscribers: 2
/TwistedBacteria/science-comm-edu Candidate members: 16 Subscribers: 35
/utne_altwire/science2 Candidate members: 16 Subscribers: 3
/physicus/scicom Candidate members: 15 Subscribers: 4
/TheUrbanOctopus/science Candidate members: 15 Subscribers: 7

Or we can order the lists with more than P subscribers by subscriber count:

/sciencebase/scientwitters 11 434
/sciencebase/sci-comms 14 142
/tlists/science-995 9 77
/teobesta/science-part-1 14 38
/TwistedBacteria/science-comm-edu 16 35
/mjrobbins/women-science-tweeps-3 9 28
/mocost/london 17 25
/emble64/science-chat 13 25
/NicolasLoubet/science-tech-world 9 23
/nemski/science 9 22
/LouWoodley/solo10-attendees 12 20
/kristinalford/sciencecomms 12 20
/OpheliaBottom/journos-and-scientists 9 20
/pfanderson/science-research 12 20
/sc_k/bloggers-writers-journali 11 19
/AndreiaASoares/sci-comm 10 19
/soloconf/solo11-attendees 10 18
/cptyler/science 20 15

Or we could order lists according to some function of both subscriber count and number of list members who are also on the original list…

The second recipe attempts to identify notable individuals in a topic area based on a set of lists in the topic area. Given a list of lists (candidate lists), such as either of the above, we can then find the members of each list, and count the number of candidate lists they appear on…

['alicebell', 17, 734, 5353, 11534]
['edyong209', 16, 735, 10504, 27207]
['skyponderer', 15, 1359, 2544, 7664]
['JoBrodie', 14, 969, 1794, 25889]
['ProfBrianCox', 14, 94, 390506, 5345]
['SciencePunk', 14, 148, 5066, 8728]
['imascientist', 14, 806, 3396, 12833]
['alokjha', 14, 880, 5123, 3774]
['sciencebase', 14, 1776, 11431, 9092]
['RogerHighfield', 14, 432, 5674, 10770]

Here, then, we have view over people who appear to be heavily represented in lists in a particular subject area, where those lists are identified by means of similarity with an initial list.

Another approach to discovering notable individuals in an area would be to take all the members of a set of candidate lists and plot their social connections, both within members of set of set of candidate lists (i.e. look to see how the members of a set of candidate lists follow each other), as well as across their whole friends lists (i.e. plot the connections between members of the candidate lists and all their friends, whether or not they appear on one or more of the candidiate lists).

PS Rather than seeding the recipe with an initially prepared list, we might also seed it with a search around a particular search term. So for example, it’s possible to generate an initial set of candidate members from a simple Twitter search. A search on a given keyword identifies a set if users who have tweeted using that keyword, which we can then use to identify a set of lists in the area and a set of users who appear across those lists. Searching on the keyword “education” turns up the following people spread across lists that mentioned people who had recently tweeted the keyword “education”:

'JobsDirectUSA', 'coolcatteacher', 'AmericaHires','educationweek', 'edutopia', 'DellEDU', 'web20classroom', 'elearnnet', 'C4LPT', 'UPSjobs'

(I suspect the original search turned up users who were represented on job related lists!)

It’s not brilliant, but it’s a start…

A tease (and reminder to myself) of something I was messing around with for a few minutes today: a List Intelligence routine that grabs all the lists the people I follow on Twitter are on and that then plots the digraph connecting my Twitter friends with those lists. Using a force directed layout of a sort, sizing nodes (non-linearly) by degree, and then filtering out friends to leave lists, this is the sort of thing I came up with:

What this approach does (sort of) is identify groups of people I follow that other people have grouped/curated onto particular lists. I suppose the next thing I need to do is compare the list based clusters with social relations clusters (i.e. the extent to which folk follow each other) to see if they turn up similar clusters…?

I’m on holiday, but I can’t stop pondering (again) how to make more of an OU flavoured collection of content currently on BBC iPlayer… Whilst bookmarking a few more BBC/OU co-pro series pages just now, I spotted one series at least has had clips posted in to the (new to me) BBC Learning Zone Class Clips:

Which got me wondering: if OU does fully funded co-pros of content that ends up in the Learning Zone Class Clips area, wouldn’t it be good if the clips listings also displayed the OU logo…?

Or maybe if the OU got a mention on the actual clips pages?

After all, the OU gets a mention, and a link, on the original programme page:

And arguably, we could do more to support learning journey related actions and resources at the more detailed, class clips level?

PS Hmmm, I wonder how things like Class Clips fit into OER space???

PPS QUIck note re: bookmarked series pages; there are also occasions when the OU co-pros an occasional episode in a longer running series, as in the case of BBC CLick Radio (World Service), which runs weekly but only has occasional OU co-pro’d episodes? From a series page linking to episode pages, how would I identify the OU co-pro’d programme pages? Or would I have to ignore series pages and just bookmark/index actual co-pro’d episode pages (if they exist?)?

PPPS Ah – this looks interesting (BBC prototype): THe Programme List (“Add entire shows, series or just episodes, See which of your programmes are available today”). So I should be able to add in lists of OU/BBC co-pros, and see a view over episodes that are currently available on iPlayer. Which makes me think: could something like The Programmes List also be used to publish and view 3rd party curated collection lists, opening up “scheduling” of BBC content to all-comers?

Ever since the Open University was founded, a relationship with the BBC has provided the OU with a route to broadcast through both television and radio. Some time ago, I posted a recipe for generating a page that showed current OU programmes on iPlayer (all rotted now…). Chatting to Liam last night, I started wondering about resurrecting this service, as well as pondering how I could easily begin to build up an archive of programme IDs for OU/BBC co-pros, so that whenever the fancy took me I could go to a current and comprehensive “OU on iPlayer” page and see what OU co-pro’d content was currently available to watch again.

Unfortunately, there doesn’t seem to be an obvious feed anywhere that gives access to this information, nor a simple directory page listing OU co-pros with links even to the parent series page or series identifier on the BBC site. (This would be lovely data to have in the OU’s open linked data store;-)

What caught my attention about this feed is that it’s focussed on growing audience around live broadcasts. This is fine if you’re tweeting added value* along with the live transmission and turning the programme into an event, but in general terms? I rarely watch live television any more, but I do watch a lot of iPlayer…

(* the Twitter commentary feed can than also be turned into expert commentary subtitles/captions, of course, using Martin Hawksey’s Twitter powered iPlayer subtitles recipe..)

There is also a “what’s on” feed available from OpenLearn (via a link – autodiscovery doesn’t seem to be enabled?), but it is rather horrible and it doesn’t contain BBC programme/series IDs (and I’m not sure the linked to pages necessarily do so, either?)

So – what to do? In the short term, as far as my tinkering goes, nothing (holidays…:-) But I think with a nice feed available, we could make quite a nice little view over OU co-pro’d content currently on iPlayer, and also start to have a think about linking in expert commentary, as well as linking out to additional resources…

See also:
Augmenting OU/BBC Co-Pro Programme Data With Semantic Tags
Linked Data Without the SPARQL – OU/BBC Programmes on iPlayer [this actually provides a crude recipe for getting access to OU/BBC programmes by bookmarking co-pro'd series pages on delicious...]

PS from @liamgh: “Just noticed that Wikipedia lists both BBC & OU as production co e.g. en.wikipedia.org/wiki/The_Virtu… RH Panel readable with dbpedia.” Interesting… so we should be able to pull down some OU/BBC co-pros by a query onto DBPedia…

PPS also from Liam – a handy recipe for generating an HTML5 leanback UI for video content identified via a SPARQL query: An HTML5 Leanback TV webapp that brings SPARQL to your living room

Another quick note from yesterday’s mini-mash at Cambridge, hosted by Ed Chamberlain, and with participation from consultant Owen Stephens, Lincoln’s Paul Stainthorp and his decentralised developers, and Sussex’s Chris Keene. This idea came from the Lincoln Jerome project (I’m not sure if this has been blogged on the Jerome project blog?), and provides a way of scrubbing MARC based records to free the metadata up from license restrictions.

The recipe goes along the lines of reconciling the record for each item with openly licensed equivalents, and creating a new record for each item where data fields are populated with content that is know to be openly licensed. In part, this relies on having a common identifier. One approach that was discussed was generating hashes based on titles with punctuation removed. This feels a bit arbitrary to me…? I’d probably reduce all the letters to the same case at the very least in an attempt to normalise the things we might be trying to hash?

I wonder if Ed’s mapping of metadata ownership might also have a role to play in developing a robust laundry service? (e.g. “Ownership” of MARC-21 records and Where exactly DOES a record come from?).

We also discussed recipes where different libraries, each with their own MARC records for a work, might be compared field by field to identify differences between the ways similar items might be catalogued differently. As well as identifying records that maybe contain errors, this approach might also enhance discovery, for example through widening a set of keywords or classification indices.

One of the issues we keep returning to is why it might be interesting to release lots of open data in a given context. Being able to pivot from a resource in one context to a resource in another context is a general/weak way of answering this question, but here are a couple of more specific issues that came up in conversation:

1) having unique identifiers is key, and becomes useful when people use the same identifier, or same-as’d identifiers, to refer to the same thing;

2) we need tool support to encourage people creating metadata to start linking in to a recognised/shared identifier spaces. I wonder if there might be value in institutions starting to publish reconciliation services that can be addressed from tools like Google Refine. (For example, How to use OpenCorporates to match companies in Google Refine or Google Refine Reconciliation Service API). Note that it might make sense for reconciliation services to employ various string similarity heuristics as part of the service.

3) we still don’t have enough compelling use cases about the benefits of linked IDs, or tools that show why it’s powerful. (I think of linked identifier spaces that are rich enough to offer benefits as if they were (super)saturated solutions, where it’s easy to crystallise out interesting things…) One example I like is how Open Corporates use reconciliation to allow you to map companies names in local council accounts to specific corporate entities. In time, one can imagine mapping company directors and local council councillors onto person entities and then starting to map these councillor-corporate-contract networks out…;-)

Finally, something Owen mentioned that resonates with some of my thinking on List Intelligence: Superduping/Work Superclusters, in which we take an ISBN, look at its equivalents using ThingISBN or xISBN, and then for each of those alternatives, look at their ThingISBN/xISBN alternatives, until we reach a limit set. (cf my approaches for looking at lists a Twitter UserID is included on, looking at the other members of the same lists, then finding the other lists they are mentioned on, etc. Note in the case of Twitter lists, this doesn’t necessarily hit a limit without the use of thresholding!)