… is animation! I haven’t seen this used in any commercial sites, but Moritz Stefaner has a flash example of an animated cloud from his thesis. You should check out his other work there too, it includes some really innovative ways of displaying tags over time, like this graph showing tag usage:

His thesis title is "Visual tools for the socio-semantic web", and he really delivers 9 completely new ways of displaying the data, most of them time-based. Even better, he has interactive and animated examples online for almost all of them. Somebody needs to hire him to develop them further.

Moritz has his own discussion on the motivations and problems with animated tag clouds. For my purposes, I want to give people a way to spot changes in the importance of email topics over time. Static tag clouds are great for showing the relative importance of a large number of keywords at a glance, and animation is a way of bringing to life the rise and decline of topics in an easy to absorb way. Intuitively, a tag cloud of words in the subjects of emails would show ‘tax’ suddenly blinking larger in the US in April. On a more subtle level, you could track product names in customer support emails, and get an idea of which were taking the most resources over time. Trying to pull that same information from the data arranged as a line graph is a lot harder.

There’s some practical problems with animating tag clouds. Static clouds are traditionally arranged with words abutting each other. This means when one changes size, it affects the position of all the words after it. This gives a very distracting effect. One way to avoid this is to accept some level of overlap between the words as they change size, which makes the result visually a lot more cluttered and hard to read. You can increase the average separation between terms, which cuts down the overlap, but does result in a lot sparser cloud.

I’m interested in trying out some other arrangement approaches. For example, I’m fond of the OS X dock animation model, where large icons do squeeze out their neighbors, but in a very unobtrusive way. I’m also hopeful there’s some non-flash ways to do this just with JavaScript.

Outlook Graph is a windows application I’ve been working on to explore the connections that exist in large mail stores. If you’re feeling brave, I’ve just released an alpha version:
Download OutlookGraphInstall_v002.msi.

It will examine all of the Outlook mail messages on the computer, and try to arrange them into a graph with related people close together. The frequency of contacts between two people shown by the darkness of the lines connecting them. My goal is to discover interesting patterns and groupings, as a laboratory for developing new tools based on email data.

The application sends no network traffic, and doesn’t modify any data, but since it’s still in the early stages of testing, I’d recommend using it only on fully backed-up machines. It runs a physics simulation to find a good graph, so on very large numbers of messages it may take a long time to process. I’ve been testing on 10-12,000 message mailboxes on my laptop, so I’ll be interested in hearing how it scales up beyond that.

I’ve found a new example of a good time-based visualization. Twitterverse shows a tag cloud of the last few hours, based on the world’s Twitter conversations. This is one of the things I’ll do with email, and it’s interesting to see how it works here.

There’s a lot of noise in the two-word phrases, with "just realized", "this morning", "this evening" and "this weekend" all showing up as the most common phrases. These don’t give much idea of what’s on people’s minds, but I can imagine you’d need a large stop word system to remove them, and that runs the risk of filtering out interesting phrases too.

A surprising amount of identifiable information came through, especially with single words. For example, xpunkx showed up in the chart, which looked like a user name. Googling it lead me to this twitter account, and then to Andy Lehman’s blog. It may just be a glitch of their implementation, but this would be a deal-breaker for most companies if it had been a secret codename gleaned from email messages. Of course, any visualization of common terms from recent internal emails would make a lot of executives nervous if it was widely accessible. Nobody wants to see "layoff" suddenly appear there and cause panic.

It’s also surprisingly changeable. Refreshing the one hour view every few minutes causes almost completely different sets of words to appear. Either the phrase frequency is very flat, eg the top phrases are only slightly more popular than the ones just below them, and so they’re easily displaced, or their implementation isn’t calculating the tag cloud quite in the way I’d expect.

The team at ideacode have done a really good job with Twitterverse, and there’s an interesting early sketch of their idea here. Max Kiesler, one of the authors, also has a great overview of time-based visualization on the web with some fascinating examples.

… is time. A mail store is one of the few sources of implicit data that has intrinsic time information baked in. The web has a very spotty and unreliable notion of time. In theory you should be able to tell when a page was last modified, but in practice this varies based on sites, and there’s no standard way (other than wayback) to look at the state of sites over an arbitrary period.

Once you’ve extracted keywords, it’s possible to do something basic like Google Trends. Here’s an example showing frequency of searches for Santiago and Wildfire:

A friend suggested something similar for corporate email; it would be good to get a feel for the mood of the company based on either common keywords, or some measure of positive and negative words in messages. This could be tracked over time, and whilst it would be a pretty crude measure, could be a good indicator of what people are actually up to. Similarly, pulling out the most common terms in search queries going through the company gateway would give an insight into what people are thinking and working on. There’s privacy concerns obviously, but the aggregation of data from a lot of people makes it a lot more anonymous and less invasive. Its harder to turn the beefburger back into a cow, the combined data is a lot less likely to contain identifying or embarassing information.

Similar to Google’s trends, but with more information and better presentation is Trendpedia. Here’s a comparison of Facebook, MySpace and Friendster over time:


So far, the examples have all been of fairly standard line graphs. There’s some intriguing possibilities once you start presenting discrete information on a timeline, and allowing interaction and exploration, especially with email. Here’s an example of a presidential debate transcript with that sort of interface, from Jeff Clark:

All of these show a vertical, one-dimensional slice of information as it changes over time. It’s also possible to indicate time for two-dimensional data. The simplest way is to accumulate values onto a plane over time, so you can see how much of the time a certain part was active. Here’s an example from Wired, showing how player location over time was plotted for Halo maps to help tweak the design:

What’s even more compelling is showing an animation of 2D data as it changes over time. The downside is that it’s a lot harder to implement, and I don’t know of too many examples. TwitterVision is one, but it’s not too useful. Mostly these sort of animations have to be client-side applications. For email, showing the exchange of messages over time on a graph is something that could give some interesting insights.

Thanks to Matthew Hurst for pointing me to a lot of these examples through his excellent blog.

A description of a document is really useful if you’re dealing with large amounts of information. If I’m searching through emails, or even when they’re coming in, I can usually decide whether a particular message is worth reading based on a short description. Unfortunately, creating a full human-quality description is an AI-complete problem, since it requires an understanding of an email’s meaning.

Automatic tag generation is a promising and practical way of creating a short-hand overview of a text, with a few unusual words pulled out. It’s somewhat natural, because people do seem to classify objects mentally using a handful of subject headings, even if they wouldn’t express a description that way in conversation.

If you asked someone what a particular email was about, she’d probably reply with a few complete sentences; "John was asking about the positronic generator specs. He was concerned they wouldn’t be ready in time, and asked you to give him an estimate." This sort of summary also requires a full AI, but it is possible to at least mimic the general form of this type of description, even if the content won’t be as high-quality.

The most common place you encounter this is on Google’s search results page:

The summary is generated by finding one or two sentences in the page that contain the terms you’re looking for. If there’s multiple occurences, usually the sentences earliest in the text are favored, along with ones that contain the most terms closest together. It’s not a very natural-looking summary, but it does do a good job at picking out the relevant quotations to what you’re looking for, and giving a good idea whether it’s actually talking about what you want to know.

Amazon’s statistically improbable phrases for books are an interesting approach, they try to identify combinations of words that are distinctive to a particular book. These are almost more like tags, and are found by a similar method to statistics-based automatic tagging, by spotting combinations that are frequent in a particular book, and not as common in a background of related items. They don’t act as a very good description in practice, they’re more useful as a tool for discovering distinctive content. I also discovered they’ve introduced capitalized phrases, which serve a similar purpose. That’s an intriguing hack on the English language to discover proper nouns, I may need to copy that approach.

The final, and most natural, type of summary is created by picking out key sentences from the text, and possibly shortening them. Microsoft Word’s implementation is the most widely used, and it isn’t very good. There’s also an online summarizer you can experiment with that suffers a lot of the same problems.

There are two big barriers to getting good summaries with this method. First, it’s hard to identify which bits of the document are actually important. Most methods use location, if it’s a heading or the start of a paragraph, and statistical frequency of unusual words, but these aren’t very good predictors. Even once you’ve picked the ones you want to use, here’s also very little guarantee that the sentences will make any sense when strung together outside the context of the full document. You often end up with a very confusing narrative. Even MS in their description of their auto summary tool acknowledge that at best it produces a starting point that you’ll need to edit.

Overall, for my purposes displaying something like Google or Amazon’s summaries for an email might be useful, though I’ll have to see if it’s any better than just showing the first sentence or two of a message. It doesn’t look like the other approaches to producing a more natural summary are good enough to be worth using.

I’ve been thinking hard about painful problems that email analysis could solve, and one of them is the use of a corporation’s email store to discover internal colleagues who have existing relationships with an external company or person you want to talk to. For example, if you want to sell to IBM, maybe there’s someone in your team who’s already talking to someone there. Or internally, you might want an introduction to someone in another department to discuss a problem, and it would be good to know who in your team had contacts there already.

I was discussing these thoughts with George Eberstadt, co-founder of nTag, and he pointed me to a couple of successful companies who are already mining email to do this, Spoke and Contact Networks.

Spoke are interesting because they’re entirely client-based, rather than running on an organization’s whole message store. They work by taking data from everybody who’s running their Outlook add-on, along with information pulled from publicly available sources, and feeding into their own database. You can then search that yourself to find information on people you’re interested in contacting, and people you know who’ve already been in touch with them. It’s effectively creating a global social network largely based on the email patterns of everybody who belongs.

Technically, it sounds like they’re doing some interesting things, such as exchanging special emails to confirm that two people really do know each other, but when I tried for my own name, I didn’t get any useful information. I also am surprised that companies would allow the export of their employees email relationships to a third-party. It may just be that this is happening under the radar, but it seems like the sort of thing that a lot of companies would want some safeguards on. The service encourages individual employees to install the software themselves, without any warning that they might be opening up the organization’s data to third-party analysis. I know at a lot of the companies I deal with would frown on this, to say the least.

Contact Networks seem much more focused on selling to corporations as a whole, rather than individual employees. They build a social graph pulling from several sources internal to the company including email and calendars, CRM, marketing databases, HR and billing systems. They use this to identify colleagues who know a particular individual, which is a succinct description of a ‘painful problem’ that companies would be willing to pay money to solve. They seem to have been very successful, with lots of big-name clients and they were just bought out by Thomson.

It’s good to see how well Contact Networks have done, it’s proof there’s demand for the sort of services I’m thinking of, even if they have already solved the immediate problems I was considering.