After the two longest months of my life, Molly finally returns home from her archeological dig in Israel tomorrow! IN YOUR FACE, stupid hole in the ground! I WIN!!

With apologies to Donnie Hoyle:

I don't have Photoshop, but I do have Paint.NET ;)

If you haven't had the pleasure of watching You Suck At Photoshop, you're missing out. Comic genius. Check it out starting here. Donnie is a socially awkward introvert who condescends and sarcastically berates you while explaining in hilarious fashion how to use many of Photoshop's most advanced features in a series of 5 minute tutorials. A new season just started on MyDamnChannel, so check it out.

In the latest episode, he notes that people have downloaded Little Ronnie (the sperm pictured above) to deface websites and modify existing pictures. He warns people NOT to do this, repeating "Please don't download Ronnie, please don't download Ronnie, please DON'T download Ronnie...dot com." When you go to pleasedontdownloadronnie.com, the title of the page is "You suck at listening to me", with a high-res version of Little Donnie to mess with on your own. The alt text of course reads "Don't do it. Seriously."

You Suck At Listening To Me

I filled up the car for the first time in two months this week, and it was $46 for 11.7 gallons. Wow. To make matters worse, I've been going to sleep too late and getting up too late to ride my bike this week, so I'm a double sucker for using a lot of gas and paying for U of M parking.

Zach called out of the blue after school because he was downtown and wanted to know if I could go on a bike ride. We rode on the Shepard Road trails, crossed Mendota Bridge, rode a bunch of Mississippi river trails, and ended up at his house in Eagan. Definitely a good ride. Unfortunately for me, I only got three hours of sleep studying for a test in Japanese class today, and I foolishly didn't eat breakfast or lunch, so I'm down a few lbs and feeling sore. TIME TO EAT!

Last week we had first oral interview of the semester in Japanese class. I got 10/10, but I accidentally said 'ichisen' instead of just 'sen' for the number 1000 when I saw it represented on a price tag in kanji (千). Overall it went really well and I felt a lot better prepared to interact and respond than in previous interviews.

Today we had our 2nd major lesson test of three total this semester. I think I did well again, but I had the month of November on the brain while I was reading through some dates in the translation section. When I saw 'December' printed, I accidentally translated it in hiragana as 'じゅういちがつ' instead of 'じゅうにがつ'. DANG!! Anyhow, I reviewed it pretty thoroughly with other students after class, and I think I got nearly everything else right.

Japanese is going really well, and my kanji flash cards are a huge help. A couple of other students saw me practicing with them and are ordering their own from White Rabbit Press. I've arranged the cards according to the order they will appear as we progress through our text book. As we cover new ones, I pull them off the pile and add them to my 'training deck' that I bring to school every day. I've aced every Kanji test so far, and the cards are definitely responsible.

I took a ride today on streets and trails from downtown Saint Paul to Afton State Park to watch my friend Zach finish a 50k race. It was about 46 miles round trip, but not nearly as hard as Zach's 50k ;) The heat was crazy and the hills were nuts. I was wondering why I felt so unprepared after the first 8 miles or so, and then I realized it was because I'd been going uphill almost the entire time. At the end of the 8 mile incline was a CRAZY hill where Bailey meets 494. After that, it was smooth sailing all the way to the park.

Here's a shot Zach's wife Jen took of Zach and me at the finish line. Zach ran like a fricken champ in that heat. Me, I got a crazy bad sunburn because I'm an idiot and didn't stop to buy any sunblock when I realized I was out.

One the way back, that last 8 miles was a mercifully easy downhill ride. At the top of the hill on Bailey, I got going about 40mph and didn't touch my pedals for almost four miles. I drank lots of fluids there and back (with some carb/electrolyte mix and some buffered electrolyte capsules), and I felt great when I got home. I headed over to my friends Marney and Dennis' house to play some volleyball. When the sun went down, Dennis and I slammed as many frappuccinos as we could stand and played Rock Band until 4am. It Was A Very Nice Day, even with the sunburn ;)

Language Log mentioned that Microsoft just acquired Powerset, a maker of Natural Language Processing search technologies. Satya Nadella writes about it in greater detail on the Live Search blog. I've used Powerset to search Wikipedia a few times, and it's really very good.

For comparison, I searched for the phrase "who is Shardik the bear?" (relevant to my recent reading) on Google (results) and Powerset (results). Now, the Wikipedia inbuilt search is just atrocious, and always has been. Using Google is an order-of-magnitude improvement, but Powerset is better still. The key difference on the main search page is that Google only highlights terms culled directly from your search pattern, whereas Powerset adds layers of highlighting to illustrate the semantic relevance of a given result to the search pattern:

 -VS-

In other words, it's really easy to see at-a-glance what Shardik is on Powerset just by scanning the highlights. Powerset highlights and emphasizes at different levels terms like character and protagonist, as well as the words that form semantic connections between my search terms 'Shardik', 'bear', and 'who' and these terms. It doesn't matter that the most relevant results to my question appear in the middle of the list vs. Google, because the most important phrases 'Shardik the ursine protagonist', 'gigantic cyborg bear named Shardik', and 'Shardik was created by North Central Positronics Ltd.' are highlighted in such a way as to immediately draw my attention.

This is just on the initial results page--the detail pages Powerset generates for Wikipedia are where the real magic happens. When you press the arrow next to a search result, a "mini viewer" opens so you can quickly navigate the article and see relevant passages:

When you mouseover the tiny paragraph representations like the one I've circled on the right, the viewer highlights the corresponding text on the left. Clicking on the representation immediately scrolls to that part of the article. It's a very handy and novel navigation widget. If you click on the result link, you see the full Wikipedia article with a few enhancements:

The most important features are the dynamically generated Article Outline widget on the right, and the "Explore Factz" button on the left. Pressing Explore Factz brings up the area I've outlined in red, which displays Things on this page and Actions on this page. Clicking any of those terms dynamically updates the article outline with clickable facts relevant to the thing or action. The Article Outline has two modes: Outline view and Factz view:

Powerset generates the outline view directly from the structure of the Wikipedia article, whereas it dynamically generates the factz using natural language processing. Clicking on the things or actions links also dynamically modifies the outline view to include relevant, clickable factz under the proper heading.

This is wicked cool and you need to play with it yourself to understand how much better this is than just using your browser's search feature to find data on a page. Wikipedia is a natural fit a system like Powerset, due Wikipedia's incredibly consistent use of styles and structure, but Microsoft/Powerset can undoubtedly incorporate a lot of this technology into general web search. This doesn't even seem like search--it's more like the pie-in-the-sky expert systems my CS profs yammered on about a looooooong time ago during my first year of college. I can't wait to see what else they do with it!

Language Log » Powerset bought by Microsoft