In the early 1990’s, when I was still using OS/2 as my primary operating system, I attended a day-long demonstration about the platform. Included in the slate of presentations was one about voice recognition. The presenter, an energetic woman, described how she was able to read and respond to hundreds of emails and Compuserve forum postings using dictation, or voice-to-text, software. She gave us a quick demo of the software’s ability by dictating, “You were right to write to me right now, Mr. Wright.” The software was able to correctly spell each variation of the phonetically identical right/write/Wright.
Of course, the software wasn’t able to do that for me when I loaded it on my computer at home. First, I had to train the software to my speech patterns, inflections, and rhythms. The training consisted of reading several passages of text, provided with the software, making corrections to the text on the screen as I went. After perhaps an hour’s work, the voice recognition was nearly 100% accurate. Through the training exercises the software was able to “learn” about my speaking habits, and accurately capture text I dictated.
In the years following that experience I have run across several other examples of “software that learns.” The one that comes to mind most readily are the Bayesian junk mail filters common in email clients today. By identifying mail you consider spam (or not spam) the filter learns how to sort your mail. Most of these tools are very accurate, with only a few false positives or negatives.
Our experience with Sibelius this weekend has led me to think that music notation software could benefit from some kind of training mode, or learning mode. Sibelius allows one to input music notation in three ways: computer keyboard, note-by-note (Steptime) using your MIDI input device, or dynamically (Flexitime) using your MIDI input device. They explain in their literature that playing a piece of music accurately to a metronome is difficult. That as humans we tend to vary the time of our playing ever so slightly. Flexitime attempts to allow for this by adjusting the tempo to match your playing speed. If you slow do, it slows down. Unfortunately there are several points of failure between the musician and the notation algorithm, not the least of which is latency introduced by the MIDI interface, and perhaps more latency introduced by the sound system in the computer. (That experienced musicians can, and do, play accurately to a metronome is a discussion for another posting.)
If Sibelius had a “learning mode” that provided several short pieces of music for the musician to play using Flexitime, the notation algorithm could examine the captured results and compare them to the stored standard. From this comparison the algorithm could “learn” about the latency characteristics of the computer, MIDI input device, and MIDI interface. While it might not completely eliminate the kind of notation problems we have seen, I feel it could go a long ways towards reducing them.
After a morning of continued experimentation with, and reading about, Sibelius, Sibylle discovered a posting online that talked about using plugins. Also, I received a comment to my earlier Sibelius posting, that specifically called out using the “Simplify Notation” plugin to clean up a score that had been entered using Flexitime.
It actually works.
Sibylle was able to play several measure of her composition normally, capturing the music in Sibelius. The score was littered with the now familiar sixteenth notes and rests. Selecting Plugins from the menu bar, and then the Simplify Notation plugin itself produces a sub-menu with several choices. The one that we used was called “Renotate performance.”
{{ $image := .ResourceGetMatch “renotate.JPG” }}
While we still feel that a plugin, and the extra step it entails, shouldn’t be necessary to capture a score, having the plugin does make capturing the score possible.
Recently Sibylle purchased an electronic keyboard for her studio, and last week the copy of Sibelius 5 that she had ordered arrived. Sibelius is one of the top two or three music notation software programs available and includes a note input mechanism called “Flexitime.” Flexitime captures the music you play on the keyboard (or other MIDI input device) directly on a score. Last evening we hooked everything up and gave it a try. We could not have been more frustrated or disappointed.
I had originally hooked the keyboard up to our iMac, which while somewhat aged (800MHz G4) is still serviceable. It also has Garage Band which we played around with a bit. Garage Band captures the MIDI information from the keyboard, but does not show that information in music notation form. Not what Sibylle needs to capture compositions.
Sibelius installed very quickly and easily on the iMac and, while a bit slow at times, ran fine on that machine. No drivers had to be added for the USB to MIDI dongle, and it was simple to follow the Sibelius setup instructions for adding a new MIDI input device to the Mac. More on how the notation feature worked in a minute.
Installing Sibelius on a Windows machine was slightly more involved, but not annoyingly so. The installation directions provided by Sibelius are thorough and even lightly humorous. Getting the MIDI dongle recognized was another matter entirely. Why peripheral manufacture insist on putting drivers on a CD with the device in the packaging, when they know that driver will be out of date in no time at all, is beyond me. Installing the included driver failed to activate our MIDI input. I had to click on the provided link, decipher a myriad of available drivers, download, and execute the new driver setup, in order to have Sibelius (and the computer) recognize the keyboard when it was attached.
Flexitime is the dynamic note input mechanism included in Sibelius. The software also allows for computer note input and steptime note input. Using the computer mouse and keyboard you can create musical scores - a lengthly and laborious process. Using step input involves the electronic keyboard, for the notes, and the mouse to select the note type from a palette. With step input you can actually just play the keyboard and get notes on to your score which are accurate in terms of which note, but aren’t accurate in terms of duration or other values. In other words, unless you change the type of note from quarter to eighth to whole, and so on, every note is recorded as a quarter note, regardless of how you play it. This input mechanism will be fantastic for producing scale or chord worksheets for students.
Flexitime tries to capture music as you play it. There are many options controlling how it performs, from what duration of notes it recognizes (whole, half, quarter, eighth, sixteenth, thirty-second), to whether or not the metronome adjusts to your playing speed or not. Both Sibylle and I tried each of these options, and many combinations of these options, trying to get an accurate capture of music being played.
We have not yet been successful. Even playing a whole note scale produces odd rests, and notes which are of the wrong duration. Setting the duration sensitivity seems to allow notes of the selected size or longer, when whole, half, or quarter are selected. Once you pick eighth or shorter as the duration sensitivity, Sibelius starts littering the composition with sixteenth notes, and even thirty-second notes.
Sibelius does provide an online help web site, but several searches didn’t reveal any useful information about our situation. Google searches did turn up several other people with basically the same problem as us. Some of the suggested answers included making sure that the sound card drivers were up to date (although why the playback channel should improve a MIDI input is beyond me). And one site said that it is possible to get Flexitime to work, but that it took a lot of fiddling with the available settings.
I spent a hour or so this morning cursing Windows and various driver manufacturers as I update the ThinkPad’s sound card driver, all to no avail. We were still unable to successful produce a score using Flexitime.
The Sibeluis literature suggests that it is difficult for even seasoned pianists to play accurately to a measured beat; there will be little variances, which will throw the notation algorithm off. So Flexitime tries to adjust to your style of play, speeding up or slowing down the beat as necessary so that the final composition has fidelity. Therein, I think, lies the problem. Sibylle has 36 years of piano playing experience, she can play a piece accurately. Having software “interpret” and “adjust” her playing basically ruins the composition capture. We will continue to “fiddle” with the various settings and, as soon as the phone support reopens on Tuesday, we will call for technical assistance.
The idea that what another says can’t hurt you is patently false. Words can do damage, often permanent and sometimes costly. So I have empathy for someone who is under an attack of words, particularly if the words are incorrect or false. However, I do not have much sympathy for someone who engages in a war of words and then cries “Foul!” when the other side responds in kind.
When I first started riding motorcycles I took the Motorcycle Safety Foundation (MSF) Basic Riding course. At the beginning of the first night of class, the instructor asked us if we thought a motorcycle could be ridden safely. After some discussion we thought that, yes, it could, provided you were very careful. He then pulled out a dictionary and read the definition of “safely.” It means “without risk”. He continued to say that, “Once you throw your leg over the seat of a motorcycle, you are telling the world you are willing to accept some risk.”
There are many endeavors in life that come with some risk attached. In fact, it might be easier to list the ones that are truly risk free. Starting a campaign or crusade to right some wrong, or convince the world that your viewpoint is right, by attacking another’s viewpoint, is one sure way to take on some risk. Waging your crusade on the Internet guarantees that the whole world could be part of the audience. “World Wide Web” isn’t just a catchy marketing phrase.
In fact, the Internet is the new play ground, one allowing anyone with access to a computer to sling insults, barbs, jabs, lies, taunts, and flames. Forums, mailing lists, and comment threads are often overrun by cretins hiding behind the anonymity that being online provides. Wiki-style pages, where anyone with an account can edit the contents, have become hugely contentious virtual spaces. Zealots create dozens of alternative IDs, known as sock-puppets, to make it appear like they have support. The best known Wiki, Wikipedia, is often in the news as a result of contentious factions on opposite sides of an issue, biography, or fact, trying to keep their view on top, as the most recent edit. If nothing else, Wikipedia is the best example I can think of for the need to think critically. One has to evaluate the information there, asking again and again, “what is the source?” Without an understanding of the source, of the foundation a given entry stands upon, any information taken away from Wikipedia can only be considered hearsay.
To their credit, Wikipedia does try to manage the often unruly mob of editors and contributors, but it is a Sisyphian task. Zealotry abounds there, and if you are one to take umbrage at what another has written or said, you would be well advised to not participate. As they say, wrestling with a pig results in you both getting muddy, and the pig likes it. Wrestle with virtual pigs in the form of trolls in comment threads or online forums, or sock-puppets on Wikipedia, and you merely make yourself their next target.
Recently there was an article about Wikipedia in general, and about one contributor in particular. The journalist tried, through various means, to discover the real world identity of one particular Wikipedian, going so far as to try and back track though IP addresses where he or she was located. At one point in the article, believing the mystery Wikipedian lived in a particular area of the city, the journalist mused about how to identify and confront him or her. In effect the journalist was stalking this person, albeit for the purposes of completing her article. What the journalist wanted to confront the Wikipedian about was his or her alleged stalking and bullying online, of the journalist’s sister. It seems that the mystery Wikipedian was using sock puppets and other tactics to run roughshod over other contributors. However, the journalist’s sister may well have had some sock puppets of her own. Or maybe not. You see, that is the entire problem with trying to confront this type of childish “he said, she said” behavior. It is never ending, and impossible to eradicate. What’s more, in her article she herself admits that after five unsuccessful attempts to contact the mystery Wikipedian, she would have to get “pushy.” Objecting when someone tapes a “kick me” sign to your back is one thing, taping one there yourself and then objecting is another thing entirely.
How do I know about this sordid little tale? I wrote a piece about Anonymity that garnered a comment. This comment was offensive to the journalist. The journalist responded to the comment, here on my site, with not one, but two comments of her own. She also sent me an email claiming that the original comment was defamatory. She wanted it removed from my site. She wanted to know,
So this afternoon, when the journalist called me on my cell phone, to ask if I would remove the comments that offended her, I was again put off and reluctant to comply. If she is willing to allow comments that don’t paint her in a flattering light to appear on an article published by her own employer, then why should she get to pick and choose which comments appear about her on an obscure, nerdy web site that largely pertains to the existential navel gazing of a middle-aged man? Beyond the potential defamation in the comment, her reason for having the comment removed, said to me during our brief phone conversation, is a fear the subject of her article may be stalking her. I hardly think the absence of one comment will stop him (or her) from doing what ever they want.
If she thought that she was going to write this article and “out” a Wikipedian for less than ethical behavior, be patted on the head and given a lollypop, and have that be the end of it, then I think she has an awful lot to learn about the Internet. There are any number of comments on the article critical of the journalism in the article, and any number of other sites critical of it, with one site noting, “She chronicled her meandering, unsuccessful saga in 4,275 paralyzing words.” The last comment on my posting indicates that this article was the subject of some discussion in a college journalism class. The Internet is merciless and unforgiving and not a place to engage in word games, unless you are willing to become a target yourself.
The site Expert Law, on their article about “Defamation, Libel and Slander Law” has this to say about why commencing a defamation action may be a bad idea:
The publicity that results from a defamation lawsuit can create a greater audience for the false statements than they previously enjoyed. For example, if a newspaper or news show picks up the story of the lawsuit, false accusations that were previously known to only a small number of people may suddenly become known to the entire community, nation, or even to the world. As the media is much more apt to cover a lawsuit than to cover its ultimate resolution, the net effect may be that large numbers of people hear the false allegations, but never learn how the litigation was resolved.
Another big issue is that defamation cases tend to be difficult to win, and damage awards tend to be small. As a result, it is unusual for attorneys to be willing to take defamation cases on a contingent fee basis, and the fees expended in litigating even a successful defamation action can exceed the total recovery.
Another significant concern is that, even where the statements made by the defendant are entirely false, it may not be possible for a plaintiff to prove all of the elements of defamation. Most people will respond to news that a plaintiff lost a defamation lawsuit by concluding that the allegations were true.
In other words, the plaintiff in a defamation action may be required to expend a considerable amount of money to bring the action, may experience significant negative publicity which repeats the false accusations, and if unsuccessful in the litigation may cement into the public consciousness the belief that the defamatory accusations were true. While many plaintiffs will be able to successfully prosecute defamation actions, the possible downside should be considered when deciding whether or not such litigation should be attempted. I fear that the journalist’s toothpaste is entirely out of the tube, and no amount of work will ever get it all back in again. However, the possibility of creating an even bigger mess is entirely within the realm of possibility.
USA Today has an article up about new school policies regarding the minimum grade possible. Instead of allowing scores below 50, this new “minimum 50” policy would have 50 be the lowest score possible. The reasoning given in the article is that having a score range from 0 - 59 makes getting an F six times as likely as getting any other score. As John Gruber of Daring Fireball points out in ‘Minimum 50’ Grading Polices,
The issue with grades and testing is placing emphasis on the grade, and not on the knowledge, or lack thereof, it represents. Tests should help teacher and student both, understand what material isn’t understood fully. Ideally, there would be time to return to the just-tested material and review it again. Something, I fear, that isn’t done in most circumstances.
Douglas Reeves, and the school boards rolling this minimum 50 plan out, should all receive falling grades - numerically and otherwise.
In the fifth grade my class was taken on a field trip to a neighboring junior high school where we heard the band play. After the arranged pieces were done, each instrument was introduced by the conductor, so we would start to know what they were and what they sounded like. After that performance we were all asked if we wanted to play an instrument. While hazy, my recollection is that all the boys wanted to play drums and the girls, flute.
Those who were interested, and whose parents could afford to rent or buy an instrument, started learning how to play. My mother was able to borrow from one of her friends, Mrs. Lamb, a cornet, which I played for the remainder of fifth grade and all of sixth. Midway through the sixth grade I started what would become four years of braces on my teeth and playing the cornet wasn’t fun, as its mouthpiece was small enough that I didn’t like the sensation of my lips pinched between the mouthpiece and my braces.
In seventh grade I switched to baritone, which had a significantly large mouthpiece. Heck, the instrument in its case was nearly as big as I was. Which led to its downfall. Miss Mathesson, the band leader, was unhappy about the lack of practicing we were all putting in at home, so she sent home practice sheets to be signed by our parents. Thirty minutes of practice each and every day was the requirement. Riding the bus too and from school every day with the baritone case perched in my lap was uncomfortable, and I soon talked my parents in to letting me quit band and join the rest of the non-instrument playing students in music class.
At some point in the next couple of years I talked my way into getting a guitar and lessons. The instructor was probably only a few years older than I at the time, with long hair and not much patience for a bumbling beginner. The hurdle that I could not overcome was tuning the guitar; this being long before electronic tuners were available, you had to tune the strings by ear to each other. The instructor was never able to help me understand what “in tune” sounded like. I was also not motivated to practice. Something I now understand isn’t an innate ability. Good teachers not only cover the material, they cover how to learn it as well.
The peak of my guitar “playing” was being able to fumble my way through three or four songs I learned through sheer memorization - all with the same finger picking pattern. I didn’t even know the notes I was playing, beyond a vague mention of the chords. In college one of my roommates, who had played piano for nearly a decade already, grew tired of my endless repetitions of the same few chords and volunteered to teach me music. I had no idea what that meant, and when he presented me with the music theory I once again recoiled from the work involved.
I have always wanted to play an instrument, to be able to make music. Watching Sibylle play is beautiful; seeing her hands gracefully coax such wonderful passages and sounds from the piano brings me great pleasure. I want to be able to play, I want to be able to produce music.
Sibylle has been teaching me about music, almost from the beginning of our relationship. Through her I have learned that practice isn’t something I, or anyone else, should just know how to do. Watching her teach, and having her answer my questions, has shown me how to go about learning this thing called music.
In the little that I know already I see parallels to the martial arts (posture, natural technique, breathing, tension and relaxing) and parallels to computer science (polymorphism, re-use, and functions). I am intrigued intellectually as well as musically. In my own way, with her loving assistance, I am starting to take my first steps into the world of playing music.
As seen on Kottke. Here are the 263 titles that I’ve seen. The original list is here.
For many years Non Sequitor has been one of my favorite daily comics. Today’s edition is particularly good.
{{ $image := .ResourceGetMatch “NonSequitor2008-05-12.jpeg” }}
It isn’t often that I add a new artist to my music collection, but thanks to Ted, and a generous birthday gift, I have not only added Loreena McKennitt to my collection, I’ve discovered a new favorite artist. Her music is hard to describe except to say that it beautifully combines Celtic origins with poetry and verse, and varied instrumentation.
Rating: Playing continuously in my ear buds for two days now.
{{ $image := .ResourceGetMatch “nightsFromTheAlhambra.jpg” }}
As you can see, I was a tiny baby forty-seven years ago. These images were taken 21 days after my birth, when my weight was back up to a healthy level. When I was nineteen my mother told me the story of my birth, and the events that followed. The day I was circumcised I apparently didn’t eat much, being somewhat cranky about the procedure. Each day thereafter I wasn’t strong enough to eat enough, and I steadily lost weight. From a birth-weight of roughly 6 pounds I dropped to 5 pounds. My mother realizing something was wrong called the doctor on a Friday afternoon.
He was able to give her the last appointment of the day, and immediately had her start feeding me with an eye dropper every 2 hours. Neither she, nor I, got much sleep, but I did start gaining weight. By May 28th, when these pictures were taken, I was back up to 6 pounds 8 ounces.
{{ $image := .ResourceGetMatch “Mark_with_hand_May_1961.jpg” }}