anyseg

Fast segmentation of text into tokens and sentences with their positions in the source text. The text is never changed: every token and sentence is a span of byte offsets into the original string, and tokens follow the conventions of Universal Dependencies. Every language is a Cargo feature: ru (Russian, on by default).

use anyseg::Segmenter;

let segmenter = Segmenter::new("ru").unwrap();
let text = "Ёж идёт домой";
let tokens: Vec<&str> = segmenter.tokenize(text).iter().map(|span| span.slice(text)).collect();
assert_eq!(tokens, ["Ёж", "идёт", "домой"]);

let doc = segmenter.segment("Ёж идёт домой\n\nКот спит");
assert_eq!(doc.sentence(1).span.slice(doc.text()), "Кот спит");

A Document holds the tokens, the words with the columns of CoNLL-U, the sentences, the paragraphs and their lines; Document::to_conllu and read_conllu write and read CoNLL-U, and a document read from CoNLL-U is written back as it was, with its comments, multiword tokens and empty nodes.

The feature cli builds the command anyseg, which writes the tokens, the sentences or CoNLL-U of files and the standard input: cargo install anyseg --features cli, then anyseg --lang ru --format conllu text.txt.

The Python package with the same segmenter is anyseg on PyPI; the documentation is at sergeyshk.github.io/anySeg.

License

MIT