For the past few months I’ve been an active member of Twine.com; a beta semantic web app riddled with AI to help us organize, share and discover information. The beta is still under heavy construction, but at this point in time, I’ve migrated entirely from Del.icio.us, personal wikis and similar online services and over to Twine.
There are several reasons for this, some of which I’ll detail here — and end by hinting at why it’s immediately relevant despite Twine being in an invitation-only beta.
For the sake of easy digest, we’ll begin with Twine’s features presented in bullets:
- Social bookmarking service
- Central storage for documents, images, videos and other data (from your machine or from the web)
- Media viewable inside Twine, bookmarked or uploaded. (videos, images, etc.)
- Collaborative platform with wiki-like editing/built in text editor
- User-created groups with discussion boards
- Intelligent analysis of added content (more on this below)
- A recommendation engine to help discover information & people relevant to you and your interests
Do note that this list is not complete, it’s just the most prominent features in the way I use Twine. As mentioned, I gradually strayed away from other online services as this provided everything in one place. One place being a very important part — I wanted my data accessible in one place instead of having to visit the diverse online services. And actively building a semantic web by using it is not a small benefit. It’s a vision and necessary addition to the web. (And to humanities toolshed).
What Twine is & Does
The Twine team recently created a new and public Twine Tour that covers the basic purpose and functions of Twine. Specific details are kept at minimum because they are under active development and change on a regular basis (Twine is in a “real” beta, see below).
The tour covers the basics and I recommend reading it in addition to this post. But here’s a look at one of the things that dazzled me from the start. The Bookmarklet.
If you’ve ever used an online bookmarking service you’re familiar with adding a link to your browsers toolbar, and then clicking it to bookmark the page you’re on for easy access online later. It’s the same concept here, except when you click the bookmarking link the pretty window below appears at the corner of your browser window.
The “Add to”, “Title” and “Summary” fields are shown truncated.
They auto-expand when clicked.
None of the information you see above was entered by hand. The summary contains the text specified in the article metadata generated by WordPress, the same regarding tags & title — plus Twine generates a thumbnail of the page. (This is not all Twine does, as you’ll see as we go on).
So things are auto-extracted; with the option of adding or modifying them manually (you can manually select a pic from the page if you want). All we need to do is click save.
This has proven a major productivity boost for me, as the extra information makes it much easier to find the items again.
Extraction Varies Depending on Item Type
The extraction works on all pages — but depending on the type of the item being bookmarked, extraction is handled differently. For example, bookmarking a product on Amazon (Snow Crash, in this case) results in an item inside Twine that looks like this:
The bookmarklet extracts important data (price, author, publisher, book cover, etc.) — identifying authors as a type of “person”, the book as a “book” type, etc. This makes it easier to find, view and organize. And the same goes for YouTube videos; info is extracted and you can view the video inside Twine. Again, the benefit is that with all the information extracted it becomes much easier to find the item. I never remember my bookmarks on del.icio.us and always have to browse long lists. So this has proven a golden feature.
Additionally I find being able to view all the info inside Twine a great relief. I no longer have to collect YouTube videos in a gallery and visit them there (or direct friends there), the same with books — I can now make a list of books to read, books I’ve read, etc. and view all the information on the same site. And oh, movies from IMDb too.
Automatic Parsing of Text (the technical side of things)
The field marked summary is automatically parsed by Twine’s AI systems to identify people, places and the likes. So for example when Stanley Kubrick is mentioned in the bookmarklet fields, or in the document you upload, or in the email you send into Twine — the system will analyze and identify him as a person (not as a mere keyword). This is called entity extraction and is applied to all text on Twine.
Under the hood, a person is defined in a larger ontology in relation to other “things”. Here’s an example of a very small portion of my own graph within Twine:
Some may not find the point of this clear. So to explain: Just as HTML enables computers to display data — this extra semantic information markup (RDF, OWL, etc.) enables computers to understand what the data is they’re displaying. And moreover, to understand what things are in relation to other things.
For an example, when we search for “Stanley Kubrick” on regular search engines, the words “Stanley” and “Kubrick” are usually regarded as mere keywords: a series of letters that the search engine then tries to find pages with those series of letters. But in the world of semantic web, the engines know “Stanley Kubrick” is a person. This results in a lot less irrelevant items from the search’s results.
Stanley Kubrick may not be the clearest example; pretend the search term is “Suzuki”: are we looking for motorcycles, or the person that created the motorcycles, Michio Suzuki?
(I Used Twine Twice While Writing this Entry)
Here are actual use cases that took place while I was writing this article.
- I had written the text below about Twine’s intelligence earlier in response to a journalist on Twine. I didn’t remember where I wrote it. To find the comment, I entered the word “Wikipedia” and specified that it should be of the type “comment”, created by “Hrafn Thórisson”. Comment found.
- I wrote in a comment on an online entry and wanted to keep it for keepsake; so I created a note for it on Twine, adding a few keywords. The process was similar to the above. Search, specify, found.
An Open, Semantic Garden (A bit more on the technicalities)
If you weren’t already aware, the systems I just described above are the basic semantic web concept: Encapsulating data in a new layer of machine processable information to help us search, find and organize the overwhelming and ever-growing sea of pictures, videos, text and whatever else we’re creating.
With every item created, uploaded or bookmarked, Twine gets dozens of bits of information ranging from automatically recognized people to products, buildings, etc. As Twine grows the potential to re-use the heaps of auto-mined data grows. Twine is learning how users work, what things mean.
And important to me is that these data structures follow standards for semantic web markup. And Twine, or its creators Radar Networks, will make the data open so that others can build applications that make use of it. I generated the graph above by accessing Twine’s data with a third party tool (RDFGravity).
As an example of further growth and intertwining with other parts of the web, I believe Twine uses machine learning & 300,000 taxonomic categories of the Wikipedia for reference. For example. Further collaboration with open semantic services are also on the radar.
With this kind of data in place we pave the road towards a more intelligent web. A web that knows that an actor is related to films and is a type of person, and that a person is a type of animal; opening up a myriad of possibilities for us to search in new and more effective ways. Hopefully a farewell to awkward search methods we use today. When Twine was unveiled, I wrote an article and put it this way:
If Twine delivers it means we’ll be getting [a site that allows us to see all our online data in one place], with intelligent frosting plus a nice warm cup of IQ Cappuccino. Not to forget that its success would make a splash in semantic web development, adding a yellow brick to the road leading away from the ever-less-productive methods of modern search and data organization.
Deus ex machina for content discovery? Not yet.
It’s not always a walk in the park. It’s a beta after all, and many things need fixing and are under construction. Many things remain undone. The auto-extraction doesn’t work well on all pages, for example. The interface needs serious tweaking and several other necessary features are still under development.
There’s More to Come
The “beta” tag is (unfortunately) often (mis)used as a pre-apology for potential user discomfort. This is not the case here. Radar Networks has stated that interface and functionality is being actively worked on (as I’ve witnessed), and that by “Beta” they mean an incomplete product.
Like I commented online on a negative review: Twine is not presented as a finished product. Not even close. It’s a product in the making; seeking active users and their feedback to help make improvements. But still, when Googling for something and becoming frustrated over messy results, I’ve caught myself on more than one occasion instinctively moving the mouse to specify item-type in the filter sidebar.
We’ve yet to see what the final, public product will look like. It’s months away (hovering somewhere around next summer). The future potential of its infrastructure and usage is immense. Considering the big picture: building up metadata is not an easy task; but the Twine system is, in my personal opinion, a great way to help it on its way. More so than any of the other alphas and betas of semantic web apps I’ve explored. But great things don’t happen overnight. As I mentioned above, this is the first wave of semantic web apps and I’m happy to support the effort via feature requests and constructive criticism. Especially because I find it useful already.
Having read the above: if you were given the chance — would you be interested in using Twine? If the answer is Yes, keep an eye on Think Artificial. If the answer is No, continue keeping an eye on Think Artificial.
Links & References
- Radar-Networks Unveils Twine
- Pourquoi J’ai Migré sur Twine (Et Comment d’Autres Services Sociaux Vont Mordre La Poussière) — Christophe G. Ducamp’s French translation of this article
- Best Technology Innovation, 5 Crunchies Finalists The Crunchies is a competition to recognize and celebrate the most compelling startups, internet and technology innovations of the year....