mikemahoney218 / lrtc

Implementation of low-resource text classification from Jiang et al. (2023)

lrtc: Low-resource text classification

This crate is a Rust implementation of the low-resource text classification method introduced in Jiang et al. (2023). This implementation allows you to choose from gzip, zstd, zlib, or deflate compression algorithms, at various levels of compression.

use lrtc::{CompressionAlgorithm, classify};

let training = vec!["some normal sentence".to_string(), "godzilla ate mars in June".into(),];
let training_labels = vec!["normal".to_string(), "godzilla".into(),];
let queries = vec!["another normal sentence".to_string(), "godzilla eats marshes in August".into(),];
// Using a compression level of 3, and 1 nearest neighbor:
println!("{:?}", classify(&training, &training_labels, &queries, 3i32, CompressionAlgorithm::Gzip, 1usize));

This method seems to perform decently well for relatively sparse training sets, and does not require the same amount of tuning as neural net methods.

use csv::Reader;
use lrtc::{classify, CompressionAlgorithm};
use std::fs::File;
let imdb = File::open("./data/imdb.csv").unwrap();
let mut reader = Reader::from_reader(imdb);

    let imdb = File::open("./data/imdb.csv").unwrap();
    let mut reader = Reader::from_reader(imdb);

    let mut content = Vec::with_capacity(50000);
    let mut label = Vec::with_capacity(50000);
    for record in reader.records() {
        content.push(record.as_ref().unwrap()[0].to_string());
        label.push(record.unwrap()[1].to_string());
    }

let predictions = classify(
    &content[0..1000],
    &label[0..1000],
    &content[40000..50000],
    3i32,
    CompressionAlgorithm::Zstd,
    5usize,
)

let correct = predictions
    .iter()
    .zip(label[40000..50000].to_vec().iter())
    .filter(|(a, b)| a == b)
    .count();
println!("{}", correct as f64 / 1000f64)
// 0.623

References

Zhiying Jiang, Matthew Yang, Mikhail Tsirlin, Raphael Tang, Yiqin Dai, and Jimmy Lin. 2023. “Low-Resource” Text Classification: A Parameter-Free Classification Method with Compressors. In Findings of the Association for Computational Linguistics: ACL 2023, pages 6810–6828, Toronto, Canada. Association for Computational Linguistics. https://aclanthology.org/2023.findings-acl.426