cn-nnzy/mbglot/eval - mooncakes.io

README
E
OptionsError
E
ProfileError
- ::detectAccumulator::detect
- ::feedAccumulator::feed
- ::lengthAccumulator::length
- ::newAccumulator::new
- ::rankAccumulator::rank
- ::resetAccumulator::reset
T
ChineseVariant
- ::classifyClassifier::classify
- ::namesClassifier::names
- ::newClassifier::new
- ::rankClassifier::rank
T
Detection
- ::accumulatorDetector::accumulator
- ::detectDetector::detect
- ::explainDetector::explain
- ::newDetector::new
- ::optionsDetector::options
- ::rankDetector::rank
- ::segmentsDetector::segments
T
Explanation
- ::allLang::all
- ::bcp47Lang::bcp47
- ::codeLang::code
- ::from_bcp47Lang::from_bcp47
- ::from_codeLang::from_code
- ::iso639_1Lang::iso639_1
- ::nameLang::name
- ::native_nameLang::native_name
- ::speakersLang::speakers
T
Markup
T
Mode
- ::allowOptions::allow
- ::denyOptions::deny
- ::min_lengthOptions::min_length
- ::modeOptions::mode
- ::newOptions::new
- ::strip_markupOptions::strip_markup
- ::nameProfile::name
- ::of_langProfile::of_lang
- ::parseProfile::parse
- ::saveProfile::save
- ::trainProfile::train
- ::trigramsProfile::trigrams
T
Removed
- ::allScript::all
- ::nameScript::name
T
Segment
T
Step
- ::original_offsetStripped::original_offset
- ::removedStripped::removed
- ::textStripped::text
T
TrigramScore
V
chinese_variant
V
detect
V
detect_script
V
explain
V
rank
V
reliable_min_confidence
V
reliable_min_length
V
reliable_min_margin
V
segments
V
short_reliable_min_confidence
V
short_text_length
V
strip_markup

cn-nnzy/mbglot/eval does not have a README file

LangStats

pub(all) struct LangStats {
  lang : @cn-nnzy/mbglot.Lang
Lang
  support : Int
  predicted : Int
  correct : Int
  precision : Double?
  recall : Double?
  f1 : Double?
} derive(Eq, @moonbitlang/core/debug.Debug
Debug)

Accuracy figures for one language. precision is None when no sample was predicted as this language, recall when it has no samples.

Report

source

pub struct Report {
  // private fields
}

The outcome of evaluate: how every sample was classified.

Report::accuracy

source

fn Report::accuracy(self : Report) -> Double

Share of samples predicted correctly, 0.0 for an empty report.

Report::confusion_markdown

source

fn Report::confusion_markdown(self : Report, langs : Array[@cn-nnzy/mbglot.Lang
Lang]) -> String

The confusion matrix over langs as a Markdown table: one row per expected language, one column per predicted language in langs, then a column for other languages and one for no prediction.

Report::correct

source

fn Report::correct(self : Report) -> Int

Number of samples predicted as their expected language.

Report::count

source

fn Report::count(self : Report, expected : @cn-nnzy/mbglot.Lang
Lang, predicted : @cn-nnzy/mbglot.Lang
Lang?) -> Int

How many samples of language expected were predicted as predicted (None: not identified).

Report::languages

source

fn Report::languages(self : Report) -> Array[@cn-nnzy/mbglot.Lang
Lang]

The expected languages, in order of first appearance in the samples.

Report::macro_f1

source

fn Report::macro_f1(self : Report) -> Double

Mean F1 over the expected languages, counting a language that was never predicted as 0.

Report::per_language

source

fn Report::per_language(self : Report) -> Array[LangStats]

Precision, recall and F1 for every expected language, in order of first appearance.

Report::to_markdown

source

fn Report::to_markdown(self : Report, confusions? : Int) -> String

A Markdown summary of the report: overall accuracy, a table with precision, recall and F1 per language (in order of first appearance), and the confusions most frequent mistakes.

Report::top_confusions

source

fn Report::top_confusions(self : Report, n : Int) -> Array[(@cn-nnzy/mbglot.Lang
Lang, @cn-nnzy/mbglot.Lang
Lang?, Int)]

The n most frequent mistakes as (expected, predicted, count), most frequent first; ties are ordered by expected language, then by prediction, in Lang::all() order with None last.

Report::total

source

fn Report::total(self : Report) -> Int

Number of samples.

Report::unidentified

source

fn Report::unidentified(self : Report) -> Int

Number of samples for which no language was predicted.

Sample

source

pub(all) struct Sample {
  lang : @cn-nnzy/mbglot.Lang
Lang
  text : String
} derive(Eq, @moonbitlang/core/debug.Debug
Debug)

A text with its known language.

evaluate

source

fn evaluate(samples : Array[Sample], predict : (String) -> @cn-nnzy/mbglot.Lang
Lang?) -> Report

Runs predict on every sample and tallies the results. predict returns None when it declines to identify a text; such a sample counts as wrong for accuracy and recall, but as a prediction of no language for precision.