Nasty.Operations.Classification behaviour (Nasty v0.3.0)

Behaviour for language-agnostic text classification.

This behaviour defines the interface for training and using text classifiers that can work with any language.

Summary

Types

model()

options()

training_data()

Callbacks

algorithms()

Returns supported classification algorithms.

classify(model, input, opts)

Classifies text or document using trained model.

train(training_data, opts)

Trains a classifier on labeled training data.

Types

model()

@type model() :: term()

options()

@type options() :: keyword()

training_data()

@type training_data() :: [
  {text :: String.t() | Nasty.AST.Document.t(), category :: atom()}
]

Callbacks

algorithms()

(optional)

@callback algorithms() :: [atom()]

Returns supported classification algorithms.

classify(model, input, opts)

@callback classify(
  model :: model(),
  input :: String.t() | Nasty.AST.Document.t(),
  opts :: options()
) :: {:ok, Nasty.AST.Classification.t()} | {:error, term()}

Classifies text or document using trained model.

Parameters

model - Trained classifier model
input - Text string or Document AST to classify
opts - Classification options

Returns

{:ok, classification} - Classification result with confidence
{:error, reason} - Classification error

train(training_data, opts)

@callback train(training_data :: training_data(), opts :: options()) ::
  {:ok, model()} | {:error, term()}

Trains a classifier on labeled training data.

Parameters

training_data - List of {text, category} tuples
opts - Training options

Returns

{:ok, model} - Trained model
{:error, reason} - Training error