This is one of the most amazing examples I’ve seen of the complexity of even simple organizational schemes. “Unicode Collation Algorithm (Unicode Technical Standard #10)” spells out in precise detail how to sort strings in what we might colloquially call “alphabetical order.” But it’s way, way, way more complex than that.
Unicode is an international standard for how strings of characters get represented within computing systems. For example, in the familiar ASCII encoding, the letter “A” is represented in computers by the number 65. But ASCII is too limited to encode the world’s alphabets. Unicode does the job.
As the paper says, “Collation is the general term for the process and function of determining the sorting order of strings of characters” so that, for example, users can look them up on a list. Alphabetical order is a simple form of collation.
Sorting inconsistent alphabets is, well, a problem. But let Technical Standard #10 explain the problem:
It is important to ensure that collation meets user expectations as fully as possible. For example, in the majority of Latin languages, ø sorts as an accented variant of o, meaning that most users would expect ø alongside o. However, a few languages, such as Norwegian and Danish, sort ø as a unique element after z. Sorting “Søren” after “Sylt” in a long list, as would be expected in Norwegian or Danish, will cause problems if the user expects ø as a variant of o. A user will look for “Søren” between “Sorem” and “Soret”, not see it in the selection, and assume the string is missing, confused because it was sorted in a completely different location.
Heck, some French dictionaries even sort their accents in reverse order. (See Section 1.3.)
But that’s nothing. Here’s a fairly random paragraph from further into this magnificent document (section 7.2):
In the DUCET, characters are given tertiary weights according to Table 17. The Decomposition Type is from the Unicode Character Database [UAX44]. The Case or Kana Subtype entry refers either to a case distinction or to a specific list of characters. The weights are from MIN = 2 to MAX = 1F16, excluding 7, which is not used for historical reasons.
Or from section 8.2:
Users often find asymmetric searching to be a useful option. When doing an asymmetric search, a character (or grapheme cluster) in the query that is unmarked at the secondary and/or tertiary levels will match a character in the target that is either marked or unmarked at the same levels, but a character in the query that is marked at the secondary and/or tertiary levels will only match a character in the target that is marked in the same way.
You may think I’m being snarky. I’m not at all. This document dives resolutely into the brambles and does not give up. It incidentally exposes just how complicated even the simplest of sorting tasks is when looked at in their full context, where that context is history, language, culture, and the ambiguity in which they thrive.
Date: July 27th, 2014 dw
Two percent of Harvard’s library collection circulates every year. A high percentage of the works that are checked out are the same as the books that were checked out last year. This fact can cause reflexive tsk-tsking among librarians. But — with some heavy qualifications to come — this is at it should be. The existence of a Long Tail is not a sign of failure or waste. To see this, consider what it would be like if there were no Long Tail.
Harvard’s 73 libraries have 16 million items [source]. There are 21,000 students and 2,400 faculty [source]. If we guess that half of the library items are available for check-out, which seems conservative, that would mean that 160,000 different items are checked out every year. If there were no Long Tail, then no book would be checked out more than any other. In that case, it would take the Harvard community an even fifty years before anyone would have read the same book as anyone else. And a university community in which across two generations no one has read the same book as anyone else is not a university community.
I know my assumptions are off. For example, I’m not counting books that are read in the library and not checked out. But my point remains: we want our libraries to have nice long tails. Library long tails are where culture is preserved and discovery occurs.
And, having said that, it is perfectly reasonable to work to lower the difference between the Fat Head and the Long Tail, and it is always desirable to help people to find the treasures in the Long Tail. Which means this post is arguing against a straw man: no one actually wants to get rid of the Long Tail. But I prefer to put it that this post argues against a reflex of thought I find within myself and have encountered in others. The Long Tail is a requirement for the development of culture and ideas, and at the same time, we should always help users to bring riches out of the Long Tail
There’s a terrific article by Helen Vendler in the March 24, 2014 New Republic about what can learn about Emily Dickinson by exploring her handwritten drafts. Helen is a Dickinson scholar of serious repute, and she finds revelatory significance in the words that were crossed out, replaced, or listed as alternatives, in the physical arrangement of the words on the page, etc. For example, Prof. Vendler points to the change of the line in “The Spirit” : “What customs hath the Air?” became “What function hath the Air?” She says that this change points to a more “abstract, unrevealing, even algebraic” understanding of “the future habitation of the spirit.”
Prof. Vendler’s source for many of the poems she points to is Emily Dickinson: The Gorgeous Nothings, by Marta Werner and Jen Bervin, the book she is reviewing. But she also points to the new online Dickinson collection from Amherst and Harvard. (The site was developed by the Berkman Center’s Geek Cave.)
Unfortunately, the New Republic article is not available online. I very much hope that it will be since it provides such a useful way of reading the materials in the online Dickinson collection which are themselves available under a CreativeCommons license that enables
non-commercial use without asking permission.
A friend is looking into the best way for a city to publish its codes and ordinances to make them searchable and reusable. What are the best schemas or ontologies to use?
I work in a law school library so you might think I’d know. Nope. So I asked a well-informed mailing list. Here’s what they have suggested, more or less in their own words:
Any other suggestions?
I had a chance to talk with Dan Brickley today, a semanticizer of the Web whom I greatly admire. He’s often referred to as a co-creator of FOAF, but these days he’s at Google working on Schema.org. He pointed me to the work Schema has been doing with online datasets, which I hadn’t been aware of. Very interesting.
Schema.org, as you probably know, provides a set of terms you can hide inside the HTML of your page that annotate what the visible contents are about. The major search engines — Google, Bing, Yahoo, Yandex — notice this markup and use it to provide more precise search results, and also to display results in ways that present the information more usefully. For example, if a recipe on a page is marked up with Schema.org terms, the search engine can identify the list of ingredients and let you search on them (“Please find all recipes that use butter but not garlic”) and display them in a more readable away. And of course it’s not just the search engines that can do this; any app that is looking at the HTML of a page can also read the Schema markup. There are Schema.org schemas for an ever-expanding list of types of information…and now datasets.
If you go to Schema.org/Dataset and scroll to the bottom where it says “Properties from Dataset,” you’ll see the terms you can insert into a page that talk specifically about the dataset referenced. It’s quite simple at this point, which is an advantage of Schema.org overall. But you can see some of the power of even this minimal set of terms over at Google’s experimental Schema Labs page where there are two examples.
The first example (click on the “view” button) does a specialized Google search looking for pages that have been marked up with Schema’s Dataset terms. In the search box, try “parking,” or perhaps “military.” Clicking on a return takes you to the original page that provides access to the dataset.
The second demo lets you search for databases related to education via the work done by LRMI (Learning Resource Metadata Initiative); the LRMI work has been accepted (except for the term useRightsUrl) as part of Schema.org. Click on the “view” button and you’ll be taken to a page with a search box, and a menu that lets you search the entire Web or a curated list. Choose “entire Web” and type in a search term such as “calculus.”
This is such a nice extension of Schema.org. Schema was designed initially to let computers parse information on human-readable pages (“Aha! ‘Butter’ on this page is being used as a recipe ingredient and on that page as a movie title“), but now it can be used to enable computers to pull together human-readable lists of available datasets.
I continue to be a fan of Schema because of its simplicity and pragmatism, and, because the major search engines look for Schema markup, people have a compelling reason to add markup to their pages. Obviously Schema is far from the only metadata scheme we need, nor does it pretend to be. But for fans of loose, messy, imperfect projects that actually get stuff done, Schema is a real step forward that keeps taking more steps forward.
Here’s a recipe for a Manhattan cocktail that I like. The idea of adding Kahlua came from a bartender in Philadelphia. I call it a Bogotá Manhattan because of the coffee.
You can’t tell by looking at this post that it’s marked up with Schema.org codes, unless you View Source. These codes let the search engines (and any other computer program that cares to look) recognize the meaning of the various elements. For example, the line “a splash of Kahlua” actually reads:
<span itemprop=”ingredients”>a splash of Kahlua</span>
“itemprop=ingredients” says that the visible content is an ingredient. This does not help you as a reader at all, but it means that a search engine can confidentally include this recipe when someone searches for recipes that contain Kahlua. Markup makes the Web smarter, and Schema.org is a lightweight, practical way of adding markup, with the huge incentive that the major search engines recognize Schema.
So, here goes:
A variation on the classic Manhattan — a bit less bitter, and a bit more complex.
Prep Time: 3 minutes
Yield: 1 drink
1 shot bourbon
1 shot sweet Vermouth
A few shakes of Angostura bitters
A splash of Kahlua
A smaller splash of grenadine or maraschino cherry juice
1 maraschino cherry and/or small slice of orange as garnish. Delicious garnish.
Shake together with ice. Strain and serve in a martini glass, or (my preference) violate all norms by serving in a small glass with ice.
Here’s the Schema.org markup for recipes. author url
Jeff Atwood [twitter:codinghorror] , a founder of Stackoverflow and Discourse.org — two of my favorite sites — is on a tear about tags. Here are his two tweets that started the discussion:
I am deeply ambivalent about tags as a panacea based on my experience with them at Stack Overflow/Exchange. Example: pic.twitter.com/AA3Y1NNCV9
Here’s a detweetified version of the four-part tweet I posted in reply:
Jeff’s right that tags are not a panacea, but who said they were? They’re a tool (frequently most useful when combined with an old-fashioned taxonomy), and if a tool’s not doing the job, then drop it. Or, better, fix it. Because tags are an abstract idea that exists only in particular implementations.
After all, one could with some plausibility claim that online discussions are the most overrated concept in the social media world. But still they have value. That indicates an opportunity to build a better discussion service. … which is exactly what Jeff did by building Discourse.org.
Finally, I do think it’s important — even while trying to put tags into a less over-heated perspective [do perspectives overheat??] — to remember that when first introduced in the early 2000s, tags represented an important break with an old and long tradition that used the authority to classify as a form of power. Even if tagging isn’t always useful and isn’t as widely applicable as some of us thought it would be, tagging has done the important work of telling us that we as individuals and as a loose collective now have a share of that power in our hands. That’s no small thing.
A mailing list I’m on is discussing GenderAvenger.com. Here’s the text from the home page:
Be A Gender Avenger
Don’t Accept It. Change It.
Panel of all men? Conference with no women speakers? Book of essays with no women authors? Do something, something simple: Point it out. Opportunities — sadly — abound. How could that be in 2013? They can be found among iconic institutions and in seemingly small bore infractions.
Seeing can be believing. Everywhere possible when women are unrepresented or underrepresented, a gender avenger will take note, take action or ask someone else to take action. No excuses. This effort requires speaking out even when it is uncomfortable. Try it. The outcome could make you smile or groan. Either way you will have a story to tell that could influence others.
The site does a poor job of explaining exactly what it wants by way of input and what the outcome will be, but the email you receive if you decide to sign up anyway cites a HuffPo article about the idea, encourages you to publicize male-dominated conferences, etc., and asks for your participation in a discussion about how to make the idea work.
In the course of the back and forth on the mailing list, one participant got angry about the site and quoted the dictionary definition of “avenger”:
verb (used with object), aÂ·venged, aÂ·vengÂ·ing.
1. to take vengeance or exact satisfaction for: to avenge a grave insult.
2. to take vengeance on behalf of: He avenged his brother.
This person knows that we know (and Gina Glanz, the site’s creator, knows) what the word “avenger” means. He’s not correcting a misuse, the way he might if she’d used “revenge” as a verb. So why is he telling us what he knows we all already know?
Very likely he’s saying that the way people take a word is how the word is defined in a dictionary. But since this mailing list has been together for well over a decade, and since no one on it has ever recommended violent action (it’s moderated by a pacifist), and since the language of the site itself talks about “speaking out even when it’s uncomfortable,” to think that the site or its supporters mean “vengeance” in its dictionary sense requires dropping a whole lot of context in favor of a slavish devotion to Mr. Webster. It would be perfectly reasonable to push back on the word because it carries bad connotations or because it doesn’t quite fit the intended meaning, but neither of those conversations is advanced by citing the dictionary definition of a common word. Rather, the argument is over territory beyond the sovereignty of a dictionary.
In short (or as the kids say, TL;DR), if you’re citing a definition of a word that everyone understands, you’re probably missing the point.
Hanan Cohen points me to a blog post by a MLIS student at Haifa U., named Shir, in which she discourses on the term “paradata.” Shir cites Mark Sample who in 2011 posted a talk he had given at an academic conference, Mark notes the term’s original meaning:
In the social sciences, paradata refers to data about the data collection process itself—say the date or time of a survey, or other information about how a survey was conducted.
Mark intends to give it another meaning, without claiming to have worked it out fully. :
…paradata is metadata at a threshold, or paraphrasing Genette, data that exists in a zone between metadata and not metadata. At the same time, in many cases it’s data that’s so flawed, so imperfect that it actually tells us more than compliant, well-structured metadata does.
His example is We Feel Fine, a collection of tens of thousands (or more … I can’t open the site because Amtrak blocks access to what it intuits might be intensive multimedia) of sentences that begin “I feel” from many, many blogs. We Feel Fine then displays the stats in interesting visualizations. Mark writes:
…clicking the Age visualizations tells us that 1,223 (of the most recent 1,500) feelings have no age information attached to them. Similarly, the Location visualization draws attention to the large number of blog posts that lack any metadata regarding their location.
Unlike many other massive datamining projects, say, Google’s Ngram Viewer, We Feel Fine turns its missing metadata into a new source of information. In a kind of playful return of the repressed, the missing metadata is colorfully highlighted—it becomes paradata. The null set finds representation in We Feel Fine.
So, that’s one sense of paradata. But later Mark makes it clear (I think) that We Feel Fine presents paradata in a broader sense: it is sloppy in its data collection. It strips out HTML formatting, which can contain information about the intensity or quality of the statements of feeling the project records. It’s lazy in deciding which images from a target site it captures as relevant to the statement of feeling. Yet, Mark finds great value in We Feel Fine.
His first example, where the null set is itself metadata, seems unquestionably useful. It applies to any unbounded data set. For example, that no one chose answer A on a multiple choice test is not paradata, just as the fact that no one has checked out a particular item from a library is not paradata. But that no one used the word “maybe” in an essay test is paradata, as would be the fact that no one has checked out books in Aramaic and Klingon in one bundle. Getting a zero in a metadata category is not paradata; getting a null in a category that had not been anticipated is paradata. Paradata should therefore include which metadata categories are missing from a schema. E.g., that Dublin Core does not have a field devoted to reincarnation says something about the fact that it was not developed by Tibetans.
But I don’t think that’s at the heart of what Mark means by paradata. Rather, the appearance of the null set is just one benefit of considering paradata. Indeed, I think I’d call this “implicit metadata” or “derived metadata,” not “paradata.”
The fuller sense of paradata Mark suggests — “data that exists in a zone between metadata and not metadata” — is both useful and, as he cheerfully acknowleges, “a big mess.” It immediately raises questions about the differences between paradata and pseudodata: if We Feel Fine were being sloppy without intending to be, and if it were presenting its “findings” as rigorously refined data at, say, the biennial meeting of the Society for Textual Analysis, I don’t think Mark would be happy to call it paradata.
Mark concludes his talk by pointing at four positive characteristics of the We Feel Fine site:? It’s inviting, paradata, open, and juicy. (“Juicy” means that there’s lots going on and lots to engage you.) It seems to me that the site’s only an example of paradata because of the other three. If it were a jargon-filled, pompous site making claims to academic rigor, the paradata would be pseudodata.
This isn’t an objection or a criticism. In fact, it’s the opposite. Mark’s post, which is based on a talk that he gave at the Society for Textual Analysis, is a plea for research thatis inviting, open, juicy, and is willing to acknowledge that its ideas are unfinished. Mark’s post is, of course, paradata.
On Wednesday and Thursday I went to the second LODLAM (linked open data for libraries, archives, and museums) unconference, in Montreal. I’d attended the first one in San Francisco two years ago, and this one was almost as exciting — “almost” because the first one had more of a new car smell to it. This is a sign of progress and by no means is a complaint. It’s a great conference.
But, because it was an unconference with up to eight simultaneous sessions, there was no possibility of any single human being getting a full overview. Instead, here are some overall impressions based upon my particular path through the event.
Serious progress is being made. E.g., Cornell announced it will be switching to a full LOD library implementation in the Fall. There are lots of great projects and initiatives already underway.
Some very competent tools have been developed for converting to LOD and for managing LOD implementations. The development of tools is obviously crucial.
There isn’t obvious agreement about the standard ways of doing most things. There’s innovation, re-invention, and lots of lively discussion.
Some of the most interesting and controversial discussions were about whether libraries are being too library-centric and not web-centric enough. I find this hugely complex and don’t pretend to understand all the issues. (Also, I find myself — perhaps unreasonably — flashing back to the Standards Wars in the late 1980s.) Anyway, the argument crystallized to some degree around BIBFRAME, the Library of Congress’ initiative to replace and surpass MARC. The criticism raised in a couple of sessions was that Bibframe (I find the all caps to be too shouty) represents how libraries think about data, and not how the Web thinks, so that if Bibframe gets the bib data right for libraries, Web apps may have trouble making sense of it. For example, Bibframe is creating its own vocabulary for talking about properties that other Web standards already have names for. The argument is that if you want Bibframe to make bib data widely available, it should use those other vocabularies (or, more precisely, namespaces). Kevin Ford, who leads the Bibframe initiative, responds that you can always map other vocabs onto Bibframe’s, and while Richard Wallis of OCLC is enthusiastic about the very webby Schema.org vocabulary for bib data, he believes that Bibframe definitely has a place in the ecosystem. Corey Harper and Debra Riley-Huff, on the other hand, gave strong voice to the cultural differences. (If you want to delve into the mapping question, explore the argument about whether Bibframe’s annotation framework maps to Open Annotation.)
I should add that although there were some strong disagreements about this at LODLAM, the participants seem to be genuinely respectful.
LOD remains really really hard. It is not a natural way of thinking about things. Of course, neither are old-fashioned database schemas, but schemas map better to a familiar forms-based view of the world: you fill in a form and you get a record. Linked data doesn’t even think in terms of records. Even with the new generation of tools, linked data is hard.
LOD is the future for library, archive, and museum data.
Here’s a list of brief video interviews I did at LODLAM:
Next Page »