Metadata for experiment result site

• Indexers for words and papers
• Training docs with words indexed
• Testing docs with words indexed, also indicated split of heldout words and model words
• Date and time
• git hash + diff from some master version hash
• Actual results
  • For each clustering method
    • For each test doc, output predicted words
• key/val pair for each test doc. ideally, testdoc share common keys. It will probably be in the form of <key, (testdoc, val)>

I've started writing the Javascript code. For the time being, I'm assuming the experiment records look something like the following. Probably each of these three records would be in a separate file, and maybe certain things like the git diff would also each have their own file.

{
gitHash: "Hashy",
expName: "demoA-4",
foo: "4",
predictionScore: [0.9, 0.5, 0.6, 0.1, 1.0, 0.0],
barValue: [0.81, 0.25, 0.36, 0.01, 1.0, 0.0]
}
{
gitHash: "Hashy",
expName: "demoA-10",
foo: "10",
predictionScore: [0.09, 0.05, 0.06, 0.01, 0.1, 0.0],
barValue: [0.81, 0.25, 0.36, 0.01, 1.0, 0.0]
}
{
gitHash: "Hashy",
expName: "demoB-5",
baz: "5",
predictionScore: [0.95, 0.75, 0.8, 0.55, 1.0, 0.5],
}

In this example, the code is at git commit with hash "Hashy"; experiment demoA was run with foo set to 4 and with foo set to 10, and experiment demoB was run once, with the baz parameter set to 5. There were six test papers; demoB just gives a prediction score for each paper, but demoA also gives some other value ("barValue") to each paper.