Full text of the NYT Interview

Are you the sole inventor of HyperPop?

Yes, however it has an idea lineage and a programmer. I am not a programmer so the menu (which I now refer to as the 'hyperwords menu', please use this term if you don't mind - I will update the site soon to reflect this!) was programmed by Mikhail Seliverstov in Russia, with project management by Steve Schmidt of eJobShop in California. I have never met Mikhail in person but working with him is incredible as he really understands that this is research and he programs very flexible code and doesn't bicker when I keep changing my mind based on ideas or user feedback.

What sparked the menu the way it is now includes the 'control-click' menu on the Mac (I don't use Windows at all) where you can hold down the 'control' key on the keyboard and click on a word or an object and get options for commands to be carried out. (Of course, with the hyperwords menu, you don't click and there is no underline - it is assumed that since all the words are interactive you don't need a constant reminder and since you are in reading mode and not editing mode, the constant availability of the point and show menu is not going to be in your way)

In working with Doug Engelbart on http://www.invisiblerevolution.net I learnt how much more flexible text manipulation was in his system NLS in the 60's and 70's. NLS is actually not that hard to use, even though it is command line based, but that is another story.

Doug had also been talking about a project he calls the HyperScope which he bills as a web-based-intermediary. This has heavily influenced this project. In his system, pages would be drastically changed to add much added functionality (including paragraph level addressability, something I am proud to show him how could be done on the web - he has always had this in NLS). Doug got programmers to hack away with a real budget, but in my mind, they had too much of their own agenda and tried too advanced interface methods. I did propose early versions of the hyperwords menu but it didn't seem advanced enough. Their system would allow the user to type in commands and press buttons for serious control on a floating menu. I also doubted the possibility of any system processing a wide group of web pages without messing them up (as you can see later, I was wrong, this is exactly what the hyperwords menu came to do).

Doing my masters under Harold Thimbleby at UCLIC a couple of years ago in London and spending some time discussing hypertext with Ted Nelson who has just been a source of constant illuminating inspiration, it dawned on me that we have philosophies on links and of 'connected environments' in an often vague sort of way but what about interactive text in general? What do you call it when you change the size of a word or spell check? This was an eye opener and prompted research into interactive text.

Two questions then became apparent; what text should we make interactive and how?

The 'how' question goes back to about 1996/7 when I designed this: http://www.liquidinformation.org/autoht.html (with the bottom half added much later).

The way I look at is that we should have many different ways of interacting with information in general and text specifically, and a neat 'magic wand' which you wave over text could be one neat interaction method. Point and get a menu. Don't point and nothing is there.

The 'what text' question was first answered by RSS, which allowed us to control the environment of our blogging system completely and was therefore an ideal testing environment. After a while it seemed like, for some reason, our blog was not the most widely read material on the planet, and we really should try to make other text interactive. I asked Mikhail to build another menu (we have had 4 different incarnations so far) which should work on as many web pages as possible. I did not expect him to be able to build a menu to work on almost any site, but he found an open source engine which allowed just that.

My jaw just dropped when I saw it work.

What's the use of making every word of every text, including "and" and "the" a hyperlink?

It's funny, I had asked Mikhail to exclude these kinds of words as it wasn't really worth the server doing all the work on them, maybe we could make the system a bit faster by excluding them. Later, when the system was even slower than it is now, he ran a test and found that the work the computer had to do to find out what words to exclude was the biggest time waster in the system!

So they are live because it's costly to make them inactive and because PR-wise it's nice to say all the words are live  :-)

Why are these links called hyperlinks rather than, simply, links?

Because they are not links. A link has a prescribed start and end. The hyperwords and interactive words you can issue commands on.

The mechanism is this, the user points to a words and selects a command - these two pieces of data goes to the server: the word and the command. Anything the server can programatically do with a command on a word can conceivably be done. At this point we are really just scratching the surface of a mountain of possibilities with a q-tip.

So they are not links, they are commands, grouped (currently) in two broad philosophical categories: to change the view of the information the user has on the screen and to reference other information (not through explicit links, through implicit links as Doug calls it, to see the word in its dictionary entry fx). These are early theoretical constructs though and will likely develop.

As to the question (sorry, got a bit carried away there) of the 'hyper' that's simple. Doug was a keynote speaker at the Hypertext conference last year and asked what they meant by the 'hyper' prefix. He got little dialog. I suggested to him that he talk about 'interactive text' instead and he said sure, but what do they mean by 'hyper'? Is it meaningful like 'meta' or other prefixes? He went on to say that if 'hyper' meant something, then maybe we could develop 'hyper views' and 'hyper editing' and so on.

What effect do you think this might have on reading?

No idea! However, using the system myself I see it as adding depth to the page and putting related items closer together.

The aims of the system are the twin goals of helping the user to navigate past useless information and more thoroughly digest useful information.

Hopefully the implicit-links functionality can help the user get to where the relevant information is, without anyone else having written a path to it. Hey, I have never put it like that before but that is really what it's all about.

Why is your emphasis right now on blogs?

Good question. It's for three reasons. I started blogging to have a good way for the sponsor of the project to know what I was up to.

Since we built our own blogging system, we had full control of all the data a processes and could therefore experiment very easily.

I have also had the privilege of working with Doug, Ted Nelson, Bruce Horn (a very early and inspired collaborator for me) and others and feel that it should be recorded. For example, the current state of the hyperwords menu and this article is a bit of a triumph for me, no question, so if anyone is ever interested in how it got to this point, have a look at the blog  :-)  It's like an open research and personal journal, not an old-fashioned one only to be read after I'm dead!

How did you get CNN.com to agree to be your trial balloon?

They haven't. I wish I knew someone there to talk to, so I have made it very clear that the entry to this is only via our research site and it does not interfere with CNN content. If you know someone there please tell them I would love to talk!

Would someone reading CNN.com on their own be able to stumble on the HyperPop menu?

No. The hyperwords menu is a web based intermediary. It basically works like this. The URL for CNN through the hyperwords intermediary looks like this:


The first part is all the URL to our server: http://www.liquid.org/hyper3/hp3/HP3_Menu?url= Which is followed by the URL to CNN, which could be pretty much any site on the net which doesn't use too much CSS (Cascading Style Sheets, a standard which seriously abstracts the 'markup' part of HyperText Markup Language).

In other words, if CNN would like to add a button on their page which says something like 'View as hyperwords' or (even better) 'Make liquid' then they could. Of course, to do something like that now would crash our server. Stability or speed has not been as important criteria as flexibility in our early days. To make the system faster and more stable is just a small matter of programming.

What will be the next step after people try out your new tool?

Deciding what commands should be on the menu and how they should work is definitively the next step. We only have the capability infrastructure now, we will have to decide what capabilities are useful.

Beyond that it's all about extending our capability to navigate through and digest useful information.

I have taken ownership of the problem (too much textual information) not any one specific solution and I feel that makes a difference. There will be many different ways of dealing with this problem.

Once this work is more out there and into the world - now it's all implementation and communicating this to people - it's time to start dreaming again. And talking. And building.

How will people be able to access HyperPop on the sites of their choice? Will they have to pay for it or will individual sites have to pay to have their texts hyper-linked?

In the demo section of liquidinformation.org they can now enter any URL.

How it will work business-wise I still don't know. This is basically both fundamental research and PR for the notion of more interactive text. I do not want to own this nor make money of it, though we desperately need further funding at this point. One dream is for this sort of interactivity to become as common as links themselves. Can't happen if someone owns it.

How did you choose which dictionaries you would link to? How did the Navajo code dictionary make it into the list?

That's pure Mikhail. All I asked for was for the server to search the dictionary and show whatever it found. He made it search a whole bunch!

Is it true that in the future people will be able to favor certain kinds of links _ say, links to The New York Times or The New York Public Library. How will this work?

The glossary and commands will have a cascading architecture:

The default menu will be the one we design.

The author will be able to add words to the document for appearance in a glossary, simply by adding a special line of code at the bottom of the document, and words with their glossary entries.

The website (admins) and web pages will also have the ability to have preset menus and glossaries assigned.

The user will be able to build their own menu and will be able to choose what commands should be on the menu.

In other words, you can browse The New York Times website with the New York Times menu and glossaries if they want to build one, or your own.