Projection scoreboard

The theories of presupposition projection that linglib formalizes, scored against recorded experiments. Every mark is the right-hand side of a theorem, reduced by Lean when this page was built; follow a mark to its theorem.

✓right: the prediction matches the observation ✗wrong: it does not –silent: the theory predicts nothing here, or nothing was observed

Verdicts

Rows are theories, columns the cells of each experiment. Where a theory is scored with a way of parsing the sentences, each column is one such way, with the number of sentences whose parse it stipulates.

The observations

As printed in the papers and recorded in Linglib/Data/Experiments.

How a mark is computed

Each step is a named declaration, so a disagreement with a verdict can be traced to the step it comes from.

  1. Fact

    A judgment as printed, with its table and page: Data.Experiments.

  2. Shared parse

    The sentence in the fragment every theory reads: Presupposition.Formula, Presupposition.QuantifiedClause.

  3. Theory's parse

    By a rule on the shared parse, or stipulated sentence by sentence, at a counted cost: Eval.Parser.

  4. Acceptance

    Which contexts accept the parse; a theory without a condition for a sentence is silent on it: Eval.Projection.Theory.

  5. Linking hypothesis

    A stated rule from acceptance to the judgment subjects give: Theory.Predicts, Chemla2009a.PredictsUniversal.

  6. Verdict

    The prediction against the observation: Eval.Verdict.

The theorems

The statements behind the marks, as their docstrings put them.