rsimmons

  const originalStringify = JSON.stringify;
  JSON.stringify = function(value) {
    if (value && value.params && value.params.profiles) {
      value.params.profiles.unshift(WEBVTT_FMT);
      // console.log('stringify', value);
    }
    return originalStringify.apply(this, arguments);
  };

  const originalParse = JSON.parse;

rsimmons

{
    "initialState": "intro",
    "states": {
        "intro": {
            "fgsound": "intro",
            "bgsound": "loop1",
            "transitions": {
                "fgend": "prompt_curious",
            }
        },

rsimmons

Ligaya feedback on salami 2015-03-25

She was already aware of the general type of thing we were working on going into this.

• On prior page, she thought at first that explanatory text was part of first difficulty group, and only applied to that. Should be more clearly separated
• She chose easiest prior setting. She knows a few Japanese words. She left default of Normal listening mode.
• Liked the sounds
• Something she did countless times was click to replay, then start typing answer. But of course input box lost focus, so she would type and then notice and have to type again. She probably did this 100 times, painful to watch.
• Some of the segments were shockingly long/difficult as context (e.g. salt, ramen, etc.). From my perspective, this was by far the biggest problem for her. As she used it more, she generally figured stuff out, but this was a constant pain/issue.
• Need to tune context length limits, or something. Better to re-use short segments than have ultra-hard long ones. Or could limit them