As the factory whistle blew and the weekend began, everybody’s favourite team of crack librarians and computational ‘experts’ found themselves coming down from a week’s worth of workshops. Or so it felt.
The first workshop was a sparsely attended affair - only Data Analyst Rachel, Data Scientist Louie and mid-table computational ‘expert’ Michael deigning to attend. In fairness, they were the only people invited. In fact, it was more of a meeting than a workshop. Still, options were discussed and drawn up, so let’s agree it was a workshop. It works better for narrative consistency.
Our dear reader is probably wondering why they met and what they were ‘workshopping’. Well, there is a certain item of software, a legacy of the reign of Data Scientist Oli. It’s used to manage information for general elections. It will probably come as little surprise to our dear reader that, by the time of the last general election, the application had left the home for aged software and entered palliative care, with poor Louie nursing it through its final moments. Clearly the software needs replacing, or future general elections will prove tricky. We already have a candidate - excuse the pun - data model for all things election related in our election results schema. What we did not have was a data model describing the workflow in the lead up to a general election, on election night itself and during the days and weeks that follow. As a result of the workshop, we do now. It’s all a bit fag-packety, but, come the new year, we hope to decorate a whiteboard or two, set up a test instance of Data Graphs, throw some data at it, and see what breaks. Fun.
The second workshop of the week was also general election preparation related, with assorted colleagues from the Library and the Parliamentary Computational Section meeting to discuss a way forward. From the Library side, Anya was there, Louie was there, and Bryn was there. From the Computational Section, we had Natalie, Tom and Mr Korris. From the liminal zone between Library and computers, Rachel and Michael also showed their faces. And as, House of Lords Ben was in attendance, we had also accidentally convened a support group of the four chairs of the Members Names Information Service User Group, past and present. The bonds run deep. The event was powered by a Tupperware full of ‘tablet,’ courtesy of a kindly librarian in Edinburgh. Thanks Librarian Jaf. Lovely stuff. It certainly added pep to proceedings.
The final workshop of the week was a little more diverse in nature, seeing colleagues from the Scottish Parliament, Senedd Cymru, Northern Ireland Assembly, and the Cabinet Office descend upon Tothill Street for a day full of adventures in the hinterlands of the Sewel Convention in general and legislative consent motions in particular. For those unfamiliar, the Sewel Convention is a UK constitutional principle whereby the UK Government agrees to ‘not normally’ legislate on devolved matters. ‘Not normally’ doing quite a bit of heavy lifting there.
Whenever the UK Government presents a public bill to the UK Parliament, the explanatory notes indicate which parts of the bill the minister believes engage devolved matters. Devolved legislatures might give the green light for Westminster to legislate - “grant consent” - or they might raise concerns or objections about devolved law being changed in this way. Such concerns - or indeed lack of concerns - are signified by the medium of legislative consent motions (LCM), these being debated and resolved in the devolved legislatures. In all of this, time is of the upmost importance, devolved legislatures needing to get their proceedings underway promptly.
For the past few years, officials in all four legislatures - and all four executives - have been attempting to keep track of where and when consent has been given or withheld. This job is made more difficult by the four legislatures relying on collaboration tools where the potential for collaboration ends at the boundaries of each organisation. Emailing copies of spreadsheets like it’s 1995. Thanks Microsoft.
Librarian Jayne and computational midfield journeyman Michael were there, called to the stage late afternoon to give a quick run through of the statutory instrument service, the treaty tracker, the procedure model that sits behind them, and the procedural maps that underpin them - including maps we’ve made with colleagues from the devolved legislatures setting out their LCM procedures in a computable fashion. Always a treat when a parliamentary clerk sees one of their procedures as a computer might see it.
Finally, Librarian Jayne demoed her first stumbling attempts at LCM data collection, via the medium of our new-ish and still somewhat shiny Procedure Browsable Space™. The collection of data is not helped by the four legislatures involved having websites that are somewhat fragmented. And fragmented in a slightly different ways. Even if we say so ourselves - those assembled did seem suitably impressed. Why not take a look. Let us know which bits you find useful. And which bits less useful. Given just a little more librarian time and a little more developer time, it’s something we’d love to continue working on. Thanks for the invite Graeme.
… for a means to populate a triplestore, then two come along at once.
For the past few months, developers in the wider Data and Search team have been struggling over how one might populate a triplestore sitting on an external Azure subscription with data from a different triplestore running inside the parliamentary network. The first attempt at synching the two stores was not as successful as we’d hoped it might have been, our motto of ‘no triple left behind’ falling slightly short. Tech Lead Andrew has taken a slightly different approach, writing new code in the shape of “the Replicator”.
Meanwhile, Jianhan has taken a different approach, concluding that two triplestores is one too many and instead updating all the things that currently POST to the internal triplestore to POST to the external triplestore instead. As of this week, he’s created a new resource group in the external subscription and installed a new triplestore in that resource group. He’s also pointed the software responsible for ingesting material from line of business systems, the indexing application used by our crack team of librarians to subject index that material, the indexing service, and the software responsible for keeping our taxonomy in sync at that new triplestore. All of this behind Azure Application Gateway and Azure API Management, with Cloudflare protection to follow shortly. Jianhan has even managed to expose a SPARQL endpoint for the new triplestore, create Workbench accounts for both Anya and Rachel, and write up a roll out and roll back plan. A frankly amazing amount of work, in little over a week. There’s a reason we keep paying the lad.
Given we now have two potential solutions to the triplestore population problem, the next job is to work out which one’s best. Data Analyst Rachel is currently working with computational super-sub Young Robert to put together a test plan for the two options. We’re told Westminster bookies are now placing bets.
In news psephological, the final table of denormalised data - depicting party performances in a boundary set across general elections - has now been removed. Michael continues to work with ‘SQL neurotic’ Rachel to add a layer of professionalism over the top of his ham-fisted efforts. As with previous denormalised data removal - weeknotes passim - this means that, when the next general election rolls round, we’ll be able to load data in a much more timely fashion. Which we’re hoping will please both elders and betters.
Last time out, we reported a new approach to the general election publication cycle, the data building over time, instead of landing in one big dump at the exact moment the rest of the world has lost interest. As a half-formed part of the plan, we’re hoping to expand our proposed four-stage solution, to a 4.5 stage solution, whereby constituency elections with verified results show the full suite of numbers, whilst those awaiting verification continue to show winners only. A new database has been compiled, the code adapted slightly and the final .5 stage is now available on a test server. Which we can’t point you at for fear of confusing Google et al. So you’ll just have to take our word for it.
Our slightly depleted Member information management team have managed to tick off one more task, Members - current and past - of the House of Commons who had UK as their place of birth in the database have been informationally shunted into England, Wales, Scotland or Northern Ireland buckets as appropriate. There remain a handful of ‘ghost profiles’ still assigned to the UK, but that is more a problem with ghost profiles than their place of birth.
In even better news, team:MNIS will not be depleted for too much longer, recent recruitment concluding successfully. Harry will be joining us in January; both Librarians Anna and Phil are already looking somewhat less frazzled.
A minor update to our beloved Egg Timer™ came about following a question from Librarian Ned. Why, he asked, did the November recesses show in calendar views as a recess in the House of Commons, but not in the House of Lords? The full explanation is really rather long, so we shan’t bore you with it here. Needless to say, it all came back - as so many Egg Timer™ related matters do - to the usual chestnut of “no account [being] taken of any time during which [both / either] House[s] are adjourned for more than four days”.
Now if we had our way, we’d say four days off is little more than a long weekend, and that the word recess should only come into play when we’re talking five days or more. But it is not our job to tell the House of Lords what is a recess and what is not. Anyway, that wee foobar is now fixed.
When parliamentary material arrives on the desks of our crack team of librarians, it is neither interlinked nor fully described. A second reading reported in Hansard, for example, does not know it was a second reading, which bill it was a second reading on, or what subjects were discussed. The job of a librarian is to take this rough hewn stone and polish it to a diamond-like gleam. Subject indexing is applied by means of a spot of subject analysis and annotation with terms from our taxonomy.
For those more familiar with taxonomies represented in SKOS, the next bit may prove surprising. Our taxonomic tooling does not - yet - use SKOS, but rather a system called Zthes. In all major senses, the two systems are similar, but subtle differences exist. And we all know what happens when subtle differences meet computers.
Unlike SKOS, non-preferred terms in Zthes are identified things. And, over the years, some of those identifiers have made their way into our triplestore as subject indexings. Now the plan, as our dear reader will well know, is to upgrade our taxonomy management software - two major versions behind what is current and no longer supported by the vendor - and the upgraded version will be SKOS-based. Which left us with a problem whereby some parliamentary material would be subject indexed with identifiers for things that no longer exist. Or do exist, but only as strings, not identified things.
Taking the bit between his teeth, Librarian Ned roped in Data Engineer Victoria and our Jianhan and set off on a tidying mission. As of this week, Ned reported we now have exactly zero uses of SES IDs corresponding to a non-preferred term. Truly excellent work all round.
Finding themselves with an hour to spare, Anya, Jayne and Michael popped online to watch a short seminar on the subject of Principles for a Digital Legislative Process: Translating Procedure into Digital. A subject rather close to their hearts. The seminar was hosted by Bússola Tech, and what could have been some god-awful vendor-centric horrorshow was instead a thoughtful tour of the current state of the art. If somewhat male dominated. The video of the session is now online. If you’re at all involved with parliamentary procedure refracted through the prism of computers, it’s well worth your time.
,
