Crate crfs 

Pure Rust implementation of Conditional Random Fields (CRF)

This library provides both training and prediction capabilities for linear-chain CRFs.

Examples

Training

use crfs::train::Trainer;
use crfs::Attribute;
use std::path::Path;

let mut trainer = Trainer::lbfgs();
trainer.verbose(true);

let xseq = vec![
    vec![Attribute::new("walk", 1.0)],
    vec![Attribute::new("shop", 1.0)],
];
let yseq = vec!["sunny", "rainy"];
trainer.append(&xseq, &yseq)?;

trainer.params_mut().set_c2(1.0)?;
trainer.train(Path::new("model.crfsuite"))?;

Prediction

use crfs::{Attribute, Model};

let model_data = std::fs::read("model.crfsuite")?;
let model = Model::new(&model_data)?;
let tagger = model.tagger()?;

let xseq = vec![
    vec![Attribute::new("walk", 1.0)],
    vec![Attribute::new("shop", 1.0)],
];
let result = tagger.tag(&xseq)?;

Re-exports

pub use self::train::Trainer;

Modules

train

Training module containing all components for training CRF models Training module for CRF models

Structs

Attribute

Tuple of attribute and its value

Model

The CRF model

Tagger

The tagger provides the functionality for predicting label sequences for input sequences using a model