Within the Context of No Context

I first read this as an essay in the New Yorker in 1980, then read the book. I’m not at all sure that I have absorbed the wisdom. From the first page: “The most powerful men were those who most effectively used the power of adult competence to enforce childish agreements.”

That is a really creepy observation from 37 years ago when applied to Donald Trump in 2017.

The New Yorker has the first page of the original essay on their website. Check it out: “Within the Context of No Context“.

This book might not be for you. It is written in a strange, oracular style. But for “television” read “the Internet” and you see this.

[The Internet] is the force of no-history, and it holds the archives of the history of no-history. […] The trivial is raised up to the place where this scale has its home; the powerful is lowered there. In the place where this scale has its home, childish agreements can be arrived at and enforced effectively—childish agreements, and agreements wearing the mask of childhood.

If that doesn’t work for you, then you can skip the book. But if it makes you want to read one more paragraph, then go for it. And dang it, it makes me want to read the whole crazy thing again.

His critique of celebrity is unsurpassed. He contrasts the various “grids” (I would say “networks”) of social connections, with gradation from close friends to people you read about in the news. Only celebrities exist in the smallest and largest grids. We make celebrities part of our close friend groups, whether they be Oprah, Ellen, Beyonce, or Tomi Lahren.

The middle distance fell away…Two grids remained. The grid of two hundred million and the grid of intimacy.

The book has been described as “Baffling, cranky, elusive, brilliant.” Not sure I could be more precise than that.

Start with Goodreads and you can get to Amazon or libraries to find Within the Context of No Context.

Building a Dummy Load

If you plan to transmit on your radio, you need a way to test your transmitter without radiating a signal. You do that by transmitting into a “dummy load”. I had a 20 Watt dummy load, but I needed one to handle 100 Watts, so I built a $40 kit in June. This was the first serious soldering I had done in years, maybe decades.

I built the Oak Hills Research RFL-100 kit. A pre-built 100 W dummy load is usually $150-200. This kit is $40.

The dummy load is twenty 5 W resistors in parallel. Here are the first two resistors, ready to be soldered.

Dummy load 1

And here we are, with half of the resistors soldered onto the board.

Dummy load 2

Here we see all the resistors soldered, the board installed in the nice enclosure, and soldered to the the input connector. I chose a BNC connector instead of the standard UHF connector. My ham shack is cabled with BNC. The enclosure is marked up under the connector mounting nuts because I ground off the paint with a Dremel tool. That was much easier than sanding the paint off.

Dummy load 3

Is it 50 Ω? Well, let’s see. Hmm 49.9 Ω is within 0.1%. My Ohmmeter is accurate to +/-0.5%, so I’ll take that as a solid 50 Ω.

Dummy load 4

Finally, let’s connect it to my new 100 Watt power amplifier on the ANT 2 port. Looks great, handling 100 Watts continuous with a 1:1 SWR.

Dummy load 5

If you have a transmitter and need a dummy load, I highly recommend building the Oak Hills Research RFL-100.

International Radio Scouting Badges

Radio Scouting is an international activity, and the World Organization of the Scout Movement (WOSM) has patches and pins for it. I love the classic design, and I also love the price, with the current exchange rate.

The WOSM Radio Scouting emblem looks a lot like the World Scout Crest, but with a radio twist. It has the same purple background, but instead of the circling rope, it has dots, like Morse Code. The fleur-de-lis is at the bottom, and the center is a globe with headphones, a neckerchief, and some lightning bolt lines. It sounds complicated, but it says “Scouts on the radio around the world” without using words.

Radio Scouting patch Radio scouting pin

Right now (late 2016), the cloth badge (patch) is $1.29 (£1.04) and the metal badge (pin) is $1.81 (£1.46). You can work out the shipping cost yourself, but I think it is worth it to wear the world-wide symbol of Radio Scouting.

The Magic of Fire: The Next Level for Campfire Cooking

There are many outdoor cookbooks, but The Magic of Fire by William Rubel is the one that makes you want to build a fire in the back yard right now and roast onions.

Let’s hear what he has to say about those roasted onions, the first recipe in the book.

The shock of high heat changes onions. Caramelized sugars combine with a hint of smoke to give them unexpected complexity. The roasting process is a sensual delight. When the charred onions are ready, spear one with a fork and hold it close to your ear. You will hear the juices churning and smell an intoxicating fragrance.

The recipe is simple. This is a trimmed version, leaving out some details (“using tongs”) and adjectives (“aromatic”).

Spread the embers. Place each onion on the embers 4-8 inches from the flames. As the outer shell begins to blister, turn the onions, several times during the roasting process. The onions are done when the outer skin is charred and the onion can be easily pierced with a knife. Aim for a cooking time of 20 to 40 minutes.

Remove the cooked onions from the fire. When cool enough to handle, cut off the bottom of each onion. The burnt outer layers will often slip off like a glove. Quarter, separating the leaves. Drizzle with olive oil, toss with herbs, and season with salt.

Ready to do that on a campout?

Roasted onions

But there is a lot more beyond that first recipe. Here are a few I’d like to try:

  • Roasted eggplant spread
  • Baked beans (needs 8-12 hours)
  • Ember-baked fish (not grilled, cooked directly on the embers)
  • Chicken in a pot (an exuberant version, with onions, heads of garlic, mushrooms, tomatoes, artichokes, and greens)
  • Brisket baked in ash
  • Pot-au-Feu (feeds 15-20!)
  • Ember-roasted vegetables
  • Ember-baked potatoes (on embers or in hot ashes)
  • Ash cakes
  • Flat bread
  • Irish soda bread
  • Grilled grapes (“As the finish to a meal, grilled grapes have no peer.”)

I checked out The Magic of Fire from our local library and I was enchanted. I think I need a copy.

It won a James Beard award. It’s available new and used on Amazon. The author has a website with even more recipes and techniques “William Rubel: Traditional Foodways”.

Measuring Search Relevance with MRR

At Chegg, we test the relevance of our search engine using customer data. We extract anonymous information about queries and clicks, then use that to automatically test improvements to search. When our search engine provides results that better match what our customers are choosing, we call that an improvement.

The most basic measurement of search quality is clickthrough rate. For each search page that is shown, how often is at least one search result clicked on? This fits well with overall website conversion, showing how search works in getting a visitor to a successful result.

We can take a more critical look at our search page by not counting all clicks equally. If we count less for clicks farther down the search result list, we can evaluate the search ranking. When visitors click on the top hit, count a full click. If they click farther down, count less than a full click. We penalize ourselves for not putting the most-desired item at the top of the list.

We do this with mean reciprocal rank (MRR). We weight each click with the reciprocal of the search result’s rank. Rank one (the top result) is scored as 1, rank 2 is scored as 1/2, rank three as 1/3, rank 4 as 1/4, and so on. Then we add those reciprocal ranks up and divide by the total to get the mean, thus mean reciprocal rank. You can also think of this as a weighted clickthrough, where clicks farther down the result list are given lower scores. MRR will always be equal to or lower than clickthrough rate.

When evaluating our search engine in test, we use clicks we have collected from the website and divide by the total number of clicks. A perfect score would be 1.0, but only if there was one correct answer for each query.

Though each customer’s search has one right answer, different customers have different right answers. The same course at different universities uses different textbooks, sometimes with the same name. Even if they use the same textbook, they may use different editions. Users who typed “financial accounting” at Chegg last winter clicked on 43 different books. That is a lot more than fit in the first page of ten hits.

Some books were more popular than others, so we can use that information to test our search engine. The most popular book should be the first result, the second most popular the second hit, and so on. Of course, we have to do this without causing problems for any other search result, like queries for “introduction to financial accounting” or “financial and managerial accounting”. Here are the top five textbooks clicked on for “financial accounting” last January, with the number of times each one was clicked on.

Click Count Book
A 145 Financial Accouting, 8th Edition, Harrison et al
B 130 Financial Accouting, 2nd Edition, Spiceland et al
C 119 Financial Accouting, 7th Edition, Weygandt et al
D 106 Financial Accouting, 6th Edition, Kimmel et al
E 80 Financial Accouting, 7th Edition, Libby et al

Now let’s calculate the MRR for the query “financial accouting”, assuming that the search engine returns the books in the ideal order: ABCDE. We multiply the number of clicks for each rank by the reciprocal rank, add them up, and divide by the total number of clicks. This is the optimal ranking, so it is the highest possible MRR value for this query.

(145 * 1 +
130 * 1/2 +
119 * 1/3 +
106 * 1/4 +
80 * 1/5) / 580 = 0.504

Real search engines are not perfect, so let’s assume our engine shows the second most popular book first, then a non-relevant book, then the rest in order. Using an ‘x’ for the non-relevant book, the results order is BxACDE.

(130 * 1 +
145 * 1/3 +
119 * 1/4 +
106 * 1/5 +
80 * 1/6) / 580 = 0.418

Now we can build a repeatable test for relevance. We collect a set of queries with click counts. A script runs each query, then scores the returned results against the click data. At the end of the run, it calculates MRR for the whole set. The final MRR is reported as an overall metric, and the MRR for each query is reported for diagnostics. The script also calculates the ideal MRR for the whole set and for each query. This can run quickly, even a few thousand queries only takes a few minutes.

The first time we run the test, the result is saved as the baseline MRR. When we change the search algorithm, we can test to see if it is better or worse than the baseline MRR. When the search algorithm is improved, we set a new, higher baseline MRR.

We can also look at individual queries where the MRR is much worse than ideal to find bugs in the search engine. When we first got our MRR testing working, I found several bugs that were easy to fix. For example, we added synonyms for “intro” and “introduction” and also made “&” synonymous with “and”.

MRR is not the only relevance metric. It is best suited when there is one correct answer. At Chegg, there is one correct textbook for each person’s search. This is a known item search or navigational search. Other metrics are better for informational searches, where partial information is gathered from multiple results. If I am planning a long weekend hike in the Sierras, the metric needs to count more than one search click, because I am gathering information from many sources. Some evaluation metrics for informational searches are nDCG, RBP, and ERR.

Regardless of which metric you choose, start gathering search and click data right now. I was surprised at how much data I needed to improve individual search queries. Twenty million clicks is a good start!

Originally published on the Chegg tech blog in December 2012.

Send your name in Morse code!

This sounds like great fun for Boy Scouts or any youth-oriented radio activity. Here is the description from Dan Romanchik’s blog (kb6nu.com) about teaching Morse at the Ann Arbor Mini Maker Faire.

As usual, I had my collection of keys and was teaching kids (and some adults) how to send their names in Morse Code. I have a chart of the characters (see right) that I tape to the table next to the key, and when someone approaches the table, I ask if they’d like to learn to send their name in Morse Code, and if they say yes, I ask them to tell me the first letter in their name.

After they tell me, I show them the character on the chart and then show them how to send it. Once they’ve successfully done that, I tell them to look up the rest of the letters and then send them as well. If they successfully do this, I thrust out my hand and say, “Nice to meet you, Joe (or whatever name they just sent me).” The look on some faces is priceless.

I love the way this goes straight to “do” with a minimum of “tell”, then gives an immediate reward. The “tell me the first letter” method is really clever and makes it almost like a magic trick. You can also manage several people at the activity, as long as you can copy very slow Morse.

Send your name in morse code

Chart in PDF and in Microsoft Word formats.

Quoted with the kind permission of KB6NU.

Update: Don mentioned two things that make it simpler for the participants. First is to use “dit” and “dah” instead of dots and dashes, to start them on sound instead of pictures. Second is to leave off the numbers, since few people have numbers in their names. I’ve updated the chart to follow his excellent advice.

Guy Shopping

A guy fashion crisis is when they stop making the shoes I’ve been wearing for the last 12 years. That is because “guy shopping” is buying the same thing in the same size as quickly as possible. But sometimes, I get wild.

My L.L. Bean field watch finally died after the most recent battery replacement. It had a good 15 or 20 years, not quite sure how long. I’ve been wearing earlier versions of the watch for decades. This one lasted longer because it had a flat mineral crystal. I’m pretty good at banging my watch against things and breaking the crystal.

This time, I decided to level up, but it took a while to find a watch that I really liked. This radical departure is a Seiko Kinetic Field Watch with a 5M82 movement. It was surprisingly hard to figure out if it had the features I wanted, like stopping the second hand (hacking), and not replacing the battery (Seiko Kinetic uses a spinning weight to charge a capacitor).

Watches

Oh, yes, I replaced the strap with a G10 NATO strap. And surprise, there is a dedicated tool for dealing with the spring bars (instead of using a screwdriver and scratching the watch case like I’ve done forever).