When you feel like the software is mostly together, is it something you would be willing to share? Sell? (I hesitate to suggest beta test. ;-)
Sure, but don't hold your breath. I have a long list of things that "mostly together" will require.
2025-07-23 21:14:41
As we've gotten around to archives...
TL;DR version: I've started playing with a toy version of the beginnings of software that will eventually be a tool for my archival project. In a couple of years it ought to be interesting :^)
I'm starting to play with toy versions of software I'll be using eventually in my role as extended family archivist building a digitized record from the hundreds/thousands of pictures and document pages that have accumulated. Everything I'm doing right now is grayscale, just so that's not a surprise to anyone who goes so far as to look at the images. Most of the images are large; you'll have to do whatever tricks your browser requires to see them at full resolution.
Text documents first. A JPEG image of a document page I snapped with my iPad is here. The original image is somewhat sharper than the one shown, since JPEG is not as good with details as Apple's HEIC format. For the time being, I use ImageMagick to convert HEICs to uncompressed grayscale.
Right now the toy assumes the document is a rectangle laying flat, and I'm taking a picture of it that's out of alignment. That makes it a linear transform problem. First step is to find the corners of the document. I'm doing something not entirely simple minded. The accuracy of the toy corner-finding code is illustrated here.
It's been a long time since I did anything with linear transforms and the matrix calculations that go with that. After some online reading to refresh my memory, and finding simple versions of code for 3x3 matrices, the toy code can do a perspective transform and produce an approximate equivalent of a 300 dot-per-inch scan (or more, or less). The page in the picture is actually a pile of several sheets, stapled, so doesn't quit meet the flat rectangle assumption. The result is shown here.
In some cases, I will want to do OCR on the images. I'm using the tesseract open-source OCR program for now. Tesseract is not a toy. When I converted the photo to an estimated 600 dpi scan and ran it through tesseract: (a) tesseract estimated the resolution as 607 dpi; (b) to a quick pass through the output, all the text in what are actual text fields are correct; and (c) flat text output is shown here.
The same toy sofware works on pictures of photographs if there's a white border so that the toy can identify corners. An approximate equivalent to a 300 dpi scan of an old Polaroid picture of my wife-to-be from before I knew her is shown here.
2025-07-22 17:53:07
At the rate software is improving, I suppose computers will be able to read to us, and write down what we say as well.
During the early 1990s I had lunch regularly with a librarian. We discussed archiving on a regular basis. Ken Burns's Civil War documentary was still pretty new. She used to say, "You want to write the source material for someone to use in 120 years to make a documentary like Burns's? Acid-free paper and pigment-based ink, my friend. And descendants willing to keep your writings in a trunk somewhere dark."
2025-07-22 17:40:21
...why and whether kids still have to learn to write by hand in our modern age.
There seems to be a consistent body of work showing that taking notes during a lecture reinforces memory, and taking notes longhand reinforces more than typing on a keyboard. That's the pseudo-academic in me speaking, of course.
For the last twelve years or so I've been using a little note-taking application that I wrote myself. There were just too many cases where pasting in an image, or having a live URL, or even just searching for a keyword seemed to justify it. Recently I've been considering going back to paper and pen.
I thought about using an Ipad with an Apple Pencil, which has gotten very much like paper and pen (so long as you don't use the eraser much*). Unfortunately, Apple has seen fit to put handwriting recognition into the OS, and insists on putting a little line under anything you write/draw that it thinks might be a date or time. I've seen many complaints about it, and people asking why Apple can't make it optional.
* One of the reasons I always took notes in ink while I was doing research work was because sometimes I wrote down something that I thought was right, and two days later discovered I was mistaken. With ink, you have to grab a different color pen and put in a dated bit with the correction.
2025-07-20 18:11:52
"Whole language" may have been a wrong turn if it really resulted in ignoring phonetics (I doubt that it really did in practice), but the fact is that proficient reading requires recognition of whole words.
To paraphrase a friend, "No one can get a PhD dissertation out of pushing phonics. You have to claim that something else is better." Or at least that something else is as/more valuable than recognizing the words early on. The new things are all well and good, but memorizing a few hundred words-as-a-chunk is still necessary. You can't sound out "bat" and "cat" forever; at some point it has to be automatic.
There are assorted postings -- the internet has made them more common -- that ask whether you can read "Aoccdrnig to rscheearch at Cmabrigde Uinervtisy, it deosn’t mttaer in waht oredr the ltteers in a wrod are, the olny iprmoetnt tihng is taht the frist and lsat ltteer be at the rghit pclae." Is this just an English thing? Does Romanized Japanese tolerate the same sort of misspellings for fluent readers?
New Math was the same sort of thing. It pushed a much broader view of what math was than just the algorithms. Look, long division is done the way it is because hundreds of years of experience informs us that it's the best way to get the right answers when you have to do a hundred division problems a day, day after day. New Math failed when the teachers pushed the broader view but didn't teach the mechanics.
2025-07-20 11:49:11
It's just convention that one does not write railwaystation or particleaccelerator or internalcombustionengine but imo those are perceived as units.
German also has the useful convention of capitalizing nouns. Using Internalcombustionengine would at least signal it's a noun, even though it starts off with an adjective. Camel notation* from computer programming would possibly be better: InternalCompustionEngine.
The example everyone remembers from college German is Handschuh (hand shoe) as a generic glove/mitten term, Fingerhandschuh for gloves, Fausthandschuh for mittens, Panzerhandschuh for armor, etc. The German is at least consistent. In English, it's another of the English/Norman dualities. Glove is from Old English; mitten is from Norman for mitten; gauntlet is from the Norman for glove. I understand there are more types of Handschuh that correspond to some of the other uses English has piled on gauntlet, like "throw down the gauntlet" or "run the gauntlet".
I'll just go ahead and invite Hartmut to explain how wrong I am :^)
* Off and on for a half-century now I have occasionally tried to adopt camel notation when I'm writing code. It never lasts, and I always go back to underscores: source_index rather than sourceIndex in something I've been writing this week.
2025-07-19 10:03:03
I have the suspicion that English will eventually end up as the world language.
More than 30 years ago now, I spent some time working with an engineering team at Ericsson, the Swedish telecom company. Ericsson's internal organization at the time had hardware being done in Sweden, operating system being done in the UK, and application software being done in Spain. By decree, the official technical language inside the company was English.
The official rules for international fencing are written in French. This leads to occasional interesting difficulties. Epee rules intentionally allow some amount of incidental body-to-body contact, but not too much. There was a great deal of debate at the FIE over how to translate the French phrase for what was not allowed to English. They finally settled on "excessive jostling".
When you feel like the software is mostly together, is it something you would be willing to share? Sell? (I hesitate to suggest beta test. ;-)
Sure, but don't hold your breath. I have a long list of things that "mostly together" will require.
As we've gotten around to archives...
TL;DR version: I've started playing with a toy version of the beginnings of software that will eventually be a tool for my archival project. In a couple of years it ought to be interesting :^)
I'm starting to play with toy versions of software I'll be using eventually in my role as extended family archivist building a digitized record from the hundreds/thousands of pictures and document pages that have accumulated. Everything I'm doing right now is grayscale, just so that's not a surprise to anyone who goes so far as to look at the images. Most of the images are large; you'll have to do whatever tricks your browser requires to see them at full resolution.
Text documents first. A JPEG image of a document page I snapped with my iPad is here. The original image is somewhat sharper than the one shown, since JPEG is not as good with details as Apple's HEIC format. For the time being, I use ImageMagick to convert HEICs to uncompressed grayscale.
Right now the toy assumes the document is a rectangle laying flat, and I'm taking a picture of it that's out of alignment. That makes it a linear transform problem. First step is to find the corners of the document. I'm doing something not entirely simple minded. The accuracy of the toy corner-finding code is illustrated here.
It's been a long time since I did anything with linear transforms and the matrix calculations that go with that. After some online reading to refresh my memory, and finding simple versions of code for 3x3 matrices, the toy code can do a perspective transform and produce an approximate equivalent of a 300 dot-per-inch scan (or more, or less). The page in the picture is actually a pile of several sheets, stapled, so doesn't quit meet the flat rectangle assumption. The result is shown here.
In some cases, I will want to do OCR on the images. I'm using the tesseract open-source OCR program for now. Tesseract is not a toy. When I converted the photo to an estimated 600 dpi scan and ran it through tesseract: (a) tesseract estimated the resolution as 607 dpi; (b) to a quick pass through the output, all the text in what are actual text fields are correct; and (c) flat text output is shown here.
The same toy sofware works on pictures of photographs if there's a white border so that the toy can identify corners. An approximate equivalent to a 300 dpi scan of an old Polaroid picture of my wife-to-be from before I knew her is shown here.
At the rate software is improving, I suppose computers will be able to read to us, and write down what we say as well.
During the early 1990s I had lunch regularly with a librarian. We discussed archiving on a regular basis. Ken Burns's Civil War documentary was still pretty new. She used to say, "You want to write the source material for someone to use in 120 years to make a documentary like Burns's? Acid-free paper and pigment-based ink, my friend. And descendants willing to keep your writings in a trunk somewhere dark."
...why and whether kids still have to learn to write by hand in our modern age.
There seems to be a consistent body of work showing that taking notes during a lecture reinforces memory, and taking notes longhand reinforces more than typing on a keyboard. That's the pseudo-academic in me speaking, of course.
For the last twelve years or so I've been using a little note-taking application that I wrote myself. There were just too many cases where pasting in an image, or having a live URL, or even just searching for a keyword seemed to justify it. Recently I've been considering going back to paper and pen.
I thought about using an Ipad with an Apple Pencil, which has gotten very much like paper and pen (so long as you don't use the eraser much*). Unfortunately, Apple has seen fit to put handwriting recognition into the OS, and insists on putting a little line under anything you write/draw that it thinks might be a date or time. I've seen many complaints about it, and people asking why Apple can't make it optional.
* One of the reasons I always took notes in ink while I was doing research work was because sometimes I wrote down something that I thought was right, and two days later discovered I was mistaken. With ink, you have to grab a different color pen and put in a dated bit with the correction.
"Whole language" may have been a wrong turn if it really resulted in ignoring phonetics (I doubt that it really did in practice), but the fact is that proficient reading requires recognition of whole words.
To paraphrase a friend, "No one can get a PhD dissertation out of pushing phonics. You have to claim that something else is better." Or at least that something else is as/more valuable than recognizing the words early on. The new things are all well and good, but memorizing a few hundred words-as-a-chunk is still necessary. You can't sound out "bat" and "cat" forever; at some point it has to be automatic.
There are assorted postings -- the internet has made them more common -- that ask whether you can read "Aoccdrnig to rscheearch at Cmabrigde Uinervtisy, it deosn’t mttaer in waht oredr the ltteers in a wrod are, the olny iprmoetnt tihng is taht the frist and lsat ltteer be at the rghit pclae." Is this just an English thing? Does Romanized Japanese tolerate the same sort of misspellings for fluent readers?
New Math was the same sort of thing. It pushed a much broader view of what math was than just the algorithms. Look, long division is done the way it is because hundreds of years of experience informs us that it's the best way to get the right answers when you have to do a hundred division problems a day, day after day. New Math failed when the teachers pushed the broader view but didn't teach the mechanics.
It's just convention that one does not write railwaystation or particleaccelerator or internalcombustionengine but imo those are perceived as units.
German also has the useful convention of capitalizing nouns. Using Internalcombustionengine would at least signal it's a noun, even though it starts off with an adjective. Camel notation* from computer programming would possibly be better: InternalCompustionEngine.
The example everyone remembers from college German is Handschuh (hand shoe) as a generic glove/mitten term, Fingerhandschuh for gloves, Fausthandschuh for mittens, Panzerhandschuh for armor, etc. The German is at least consistent. In English, it's another of the English/Norman dualities. Glove is from Old English; mitten is from Norman for mitten; gauntlet is from the Norman for glove. I understand there are more types of Handschuh that correspond to some of the other uses English has piled on gauntlet, like "throw down the gauntlet" or "run the gauntlet".
I'll just go ahead and invite Hartmut to explain how wrong I am :^)
* Off and on for a half-century now I have occasionally tried to adopt camel notation when I'm writing code. It never lasts, and I always go back to underscores: source_index rather than sourceIndex in something I've been writing this week.
I have the suspicion that English will eventually end up as the world language.
More than 30 years ago now, I spent some time working with an engineering team at Ericsson, the Swedish telecom company. Ericsson's internal organization at the time had hardware being done in Sweden, operating system being done in the UK, and application software being done in Spain. By decree, the official technical language inside the company was English.
The official rules for international fencing are written in French. This leads to occasional interesting difficulties. Epee rules intentionally allow some amount of incidental body-to-body contact, but not too much. There was a great deal of debate at the FIE over how to translate the French phrase for what was not allowed to English. They finally settled on "excessive jostling".