sentiment-basic

Simple, deterministic positive/negative word-count sentiment scoring with built-in word lists. Pure Rust, zero dependencies, no network, no model downloads — the same lightweight signal the BeLikeNative AI writing assistant uses for a fast first-pass tone read.

What this is

A lexicon scorer, not a machine-learning model. It counts matches against two small, hand-curated lists of unambiguous positive and negative words and reports the balance. That makes it:

• fast — microsecond-scale, suitable for hot paths and huge corpora;
• fully deterministic — identical input always yields identical output;
• auditable — the entire "model" is two &[&str] constants you can read;
• offline — no API keys, no network, no runtime downloads.

The trade-off is that it does not understand negation ("not good"), sarcasm, intensifiers, or context. For nuanced analysis, pair it with a real model. For a stable, inspectable baseline — word counts, dashboards, gating, A/B signal — it is exactly right.

Install

[dependencies]
sentiment-basic = "0.1"

Quick start

use sentiment_basic::score;

let r = score("I love this bright happy day. It is awful and terrible though.");
println!("label:    {}", r.label());        // positive | negative | neutral
println!("net:      {}", r.net());          // i64, +positive - negative
println!("compare:  {:.2}", r.comparison()); // -1.0..=1.0
println!("pos share:{:.2}", r.positive_share()); // 0.0..=1.0

API

score(&str) returns a Sentiment:

Matching is case-insensitive, whole-word, and punctuation-safe. The lexicons are exported as POSITIVE_WORDS and NEGATIVE_WORDS if you want to extend or inspect them.

Determinism & safety

No randomness, no global state, no unsafe. Safe to call from multiple threads. Empty input returns a zero-hit "neutral" result.

License

MIT. Built for BeLikeNative.