Crate nerrs 

nerrs

CRF-based Named Entity Recognition for Persian text.

Uses crfrs for model training and Viterbi inference.

Entity types

The default tagset uses BIO encoding over these entity classes:

Tag	Meaning
O	Outside any entity
B-PER / I-PER	Person name
B-ORG / I-ORG	Organisation
B-LOC / I-LOC	Location
B-DAT / I-DAT	Date
B-TIM / I-TIM	Time
B-MON / I-MON	Money / currency amount
B-PCT / I-PCT	Percentage
B-EVE / I-EVE	Event

Quick start

use nerrs::NerTagger;

let mut tagger = NerTagger::new();
tagger.load_model("ner.model").unwrap();
let entities = tagger.tag(&["علی", "به", "تهران", "رفت", "."]).unwrap();
// → [("علی","B-PER"), ("به","O"), ("تهران","B-LOC"), ("رفت","O"), (".","O")]

Training

use nerrs::{NerTagger, crfrs::TrainConfig};

// IOB-tagged corpus: Vec<Vec<(word, ner_tag)>>
let corpus: Vec<Vec<(String, String)>> = vec![];
NerTagger::train_and_save(&corpus, "ner.model", TrainConfig::default()).unwrap();

Re-exports

pub use error::Error;

pub use error::Result;

pub use tagger::extract_entities;

pub use tagger::NerTagger;

pub use crfrs;

Modules

error

Error types for nerrs.

features

Feature extraction for Persian NER.

tagger

CRF-based Persian NER tagger.

Constants

DEFAULT_LABELS

Default NER labels in BIO encoding.