Package org.apache.tika.ml

Interface FeatureExtractor<T>

Type Parameters:

T - the raw input type (e.g. String for text, byte[] for raw bytes)

All Known Implementing Classes:

Utf16ColumnFeatureExtractor

public interface FeatureExtractor<T>

Generic feature extractor that maps an input of type T to a fixed-length integer feature vector suitable for a LinearModel.

Method Summary

Modifier and Type	Method	Description
int[]	extract(T input)	Extract features from the given input.
default int	extractSparseInto(T input, int[] dense, int[] touched)	Sparse extraction into caller-owned reusable buffers: populates dense with feature counts, writes the indices of non-zero entries into touched, and returns how many indices were written.
int	getNumBuckets()

Method Details

extract

int[] extract(T input)

Extract features from the given input.

Parameters:

input - raw input (may be null)

Returns:

int array of length getNumBuckets() with feature counts

getNumBuckets

int getNumBuckets()

Returns:

number of hash buckets (feature-vector dimension)

extractSparseInto

default int extractSparseInto(T input,
 int[] dense,
 int[] touched)

Sparse extraction into caller-owned reusable buffers: populates dense with feature counts, writes the indices of non-zero entries into touched, and returns how many indices were written. Callers are responsible for clearing the touched entries of dense before reuse.

Default implementation delegates to extract(T). Extractors that can do better (avoid allocating the full dense vector, or scan the input only once) should override.