Tag Archive: voice recognition

Day 89: Victory

I cannot begin to describe how happy this makes me:

Terminal output from Xcode voice recognition plugin

And this:

Xcode voice recognition plugin

I went back to the voice recognition plugin. I thought about it differently. And it works.

It’s not perfect. Far from it. But the key bottleneck is broken. This basically means everything I want to build, from a functionality perspective, I’ll be able to build.

I’ve never actually experienced this in coding before. I always just had someone help me through the hard parts, which was good for productivity but meant I never really learned to work through a tough problem in code. It’s awesome.

Referencing the “Day 89” in the title of this post, it’s Day 89 since leaving my job. Day 31 of Hacker School.

FacebookTwitterGoogle+Share

Things Accomplished versus Distance Covered

I’m in week four of Hacker School.

It’s awesome. But I’m struggling with a dilemma: balancing the tough problem and the maximum ground.

I view Hacker School in terms of two metrics: Things Accomplished and Distance Covered.

Of course I want to accomplish some things. I have a whole list of things I want to accomplish, some while at Hacker School and some afterward.

But the whole point of doing Hacker School, for me, is to learn the maximum possible amount in the time given, so that I become a much better programmer for whatever comes next.

This means I also need to keep an eye on the Distance Covered.

It doesn’t mean I will abandon tough problems forever; but it does mean I don’t want to spend the entire 11 weeks on one problem without any guarantee that I’ll solve it in that timeframe.

Asking the Right Question

So I’m forced to balance things. The right question is not, “How long will I spend on a particular project?”

The right question for me is, “Given the rate of progress I am making, the amount of code I am writing, and the amount of learning going on, how long will I spend on a particular project at Hacker School?”

That’s why I’ve decided to set the voice recognition project aside for now and continue forward with my Hacker School plan by turning my attention to Swift for the next two weeks.

But I Love Tough Problems

I wasn’t too happy about this decision, because my experience has been that the projects and problems I delve into tenaciously and refuse to let go until I’ve figured them out always produce my greatest successes and joy. It was true with high-speed trading risk controls at my prior job. It was true with health issues that I defeated.

But those projects take years, not weeks.

I don’t have years at Hacker School. I have time to learn a lot about a lot, meet and enjoy working with great people, and build a base for tackling the really tough problems afterward.

The Nights-and-Weekends Compromise

I discovered that the voice recognition project is 100% doable — it just requires a lot of work. More work than I have time to dedicate to it at Hacker School.

So, my compromise is to make it a nights-and-weekends project, while spending my time in the Hacker School space completing my Hacker School plan.

That means Swift for the next two weeks and then a shift to JavaScript and node.js.

Meanwhile, I’ll be mulling the voice recognition project in my spare time, which is probably the best way to tackle tough problems that are at an impasse. I will figure it out. But I will also accomplish other things and cover the distance.

Say What? My Voice Recognition Dilemma

I tried Mac Dictation this week. My goal was to get a few simple sentences to appear in a text editor. Initial experiments were not successful. The full sentences I attempted were garbled beyond recognition when they appeared on-screen, so I fell back to word-by-word communication.”Make.” Made. “Maaaaaake….” Mate. No. Try again. “Maaaaaake….” Made. No. “Create….” Create. Aha. Victory is mine! “An.” In.

I progressed word by word, slowly and carefully, and my text editor dutifully showed the words on the screen with numerous errors. This was not production-ready. No one wants to talk as if they’re scolding a recalcitrant pet.

To be clear, I’m not a marble-mouthed mumbler. I’ve recorded books on tape for Recording for the Blind and Dyslexic (now Learning Ally), and I know I can enunciate well. But Dictation couldn’t understand my voice.

I knew that if I couldn’t get text to appear on-screen, I couldn’t complete my project. I’ve started coding the next part and it’s feasible, so I really want this to work!

So, I Googled for my options:

1.) Try to improve Dictation’s performance by improving my computer setup;

2.) Buy Dragon Dictate and hope it runs on my underpowered 2011 MacBook Air and has better accuracy than Dictation.

3.) Abandon OS X entirely and do the project in a Windows virtual machine with Dragon Naturally Speaking for Windows.

4.) Try using OpenEars to build an iOS app that does the same thing.

Only the first two options really get me to my initial goal; the last two are pivots that require adapting to different platforms and approaches.

So, tomorrow I’m buying a USB headset to eliminate ambient noise and provide a clearer dictation experience. I really hope this improves Mac Dictation’s performance by leaps and bounds. If it does, I’ll be off to the races. If it doesn’t, I’ll be installing Dragon Dictate on a machine not made to handle it.

I’m heartened by the discovery that once I get the text on the screen, my goals can be accomplished. I’m discouraged by how difficult it is to actually get the text on the screen. But discouraged does not mean “gave up.” It means I’m thinking hard about the options.