Lodestar.Text 0.6.0. This page is frozen at that release. Read the current documentation for what main says now. A link to a decision or a migration page follows main, and leaves the archive.

Weight a corpus against the vocabulary and frequencies already learned.

public CsrMatrix Transform(IEnumerable<string> documents)

Parameters — documents is the corpus to weight. Terms absent from the learned vocabulary are dropped.

Returns — CsrMatrix, as wide as the fit, weighted and normalized.

Exceptions — InvalidOperationException when nothing has been fitted yet. ArgumentNullException when documents is null.

Example — the fit's width, whatever this corpus holds.

using Lodestar.Abstractions;
using Lodestar.Text.Vectorization;

var tv = new TfidfVectorizer();
tv.Fit(["the cat eats", "the dog eats", "the cat and the dog"]);

CsrMatrix weighted = tv.Transform(["the cat", "nothing here matches"]);

int width = weighted.ColumnCount;  // => 5
double empty = weighted.RowL2Norm(1);  // => 0

Remarks — the second document shares no term with the vocabulary, so its row is empty and its norm is 0 rather than 1. Normalization leaves an all-zero row alone rather than dividing by zero, so an unmatched document stays visible as a zero vector instead of becoming a NaN one — worth checking for, because a zero vector has cosine similarity 0 with everything and will quietly rank last rather than erroring.

Applies to — net10.0, netstandard2.0.

See also — TfidfVectorizer.Fit, CsrMatrix.NormalizeRows.