english is a blazing fast and light weight English inflection library written in Rust. Total bundled data size is less than 1 MB. It provides extremely accurate verb conjugation and noun/adjective declension based on highly processed Wiktionary data, making it ideal for real-time procedural text generation.
Evaluation of the English inflector (extractor/main.rs/check_*) and performance benchmarking (examples/speedmark.rs) shows:
| Part of Speech | Correct / Total | Accuracy | Throughput (calls/sec) | Time per Call |
|---|---|---|---|---|
| Nouns | 238106 / 238549 | 99.81% | 5,228,300 | 191 ns |
| Verbs | 158056 / 161643 | 97.78% | 8,473,248 | 118 ns |
| Adjectives | 119200 / 119356 | 99.86% | 11,999,052 | 83 ns |
Note: Benchmarking was done under a worst-case scenario; typical real-world usage is 50~ nanoseconds faster.
cargo add english
Then in your code:
use english::*;
fn main() {
// --- Mixed Sentence Example ---
let subject_number = Number::Plural;
let run = Verb::present_participle("run"); // running
let child = Noun::from("child").with_specifier(run); //running child
let subject = English::noun(child, &subject_number); //running children
let verb = English::verb(
"steal",
&Person::Third,
&subject_number,
&Tense::Past,
&Form::Finite,
); //stole
let object = Noun::count_with_number("potato", 7); //7 potatoes
let sentence = format!("The {} {} {}.", subject, verb, object);
assert_eq!(sentence, "The running children stole 7 potatoes.");
// --- Nouns ---
// Note that noun(), count(), etc can work on both strings and Noun struct
let jeans = Noun::from("pair").with_complement("of jeans");
assert_eq!(Noun::count_with_number(jeans, 3), "3 pairs of jeans");
// Regular plurals
assert_eq!(English::noun("cat", &Number::Plural), "cats");
// Add a number 2-9 to the end of the word to try different forms.
// Can use plural()
assert_eq!(Noun::plural("die2"), "dice");
// Use count function for better ergonomics if needed
assert_eq!(Noun::count("man", 2), "men");
// Use count_with_number function to preserve the number
assert_eq!(Noun::count_with_number("nickel", 3), "3 nickels");
// Invariant nouns
assert_eq!(English::noun("sheep", &Number::Plural), "sheep");
// --- Verbs ---
// All verb functions can use either strings or Verb struct
let pick_up = Verb::from("pick").with_particle("up");
// Helper functions: past() , third_person(), present_participle(), infinitive() etc.
assert_eq!(Verb::past(&pick_up,), "picked up");
assert_eq!(Verb::present_participle("walk"), "walking");
assert_eq!(Verb::past_participle("go"), "gone");
// Add a number 2-9 to the end of the word to try different forms.
assert_eq!(Verb::past("lie"), "lay");
assert_eq!(Verb::past("lie2"), "lied");
// "to be" has the most verb forms in english and requires using verb()
assert_eq!(
English::verb(
"be",
&Person::First,
&Number::Singular,
&Tense::Present,
&Form::Finite
),
"am"
);
// --- Adjectives ---
// Add a number 2-9 to the end of the word to try different forms. (Bad has the most forms at 3)
assert_eq!(English::adj("bad", &Degree::Comparative), "more bad");
assert_eq!(English::adj("bad", &Degree::Superlative), "most bad");
assert_eq!(Adj::comparative("bad2"), "badder");
assert_eq!(Adj::superlative("bad2"), "baddest");
assert_eq!(Adj::comparative("bad3"), "worse");
assert_eq!(Adj::superlative("bad3"), "worst");
assert_eq!(Adj::positive("bad3"), "bad");
// --- Pronouns ---
assert_eq!(
English::pronoun(
&Person::First,
&Number::Singular,
&Gender::Neuter,
&Case::PersonalPossesive
),
"my"
);
assert_eq!(
English::pronoun(
&Person::First,
&Number::Singular,
&Gender::Neuter,
&Case::Possessive
),
"mine"
);
// --- Possessives ---
assert_eq!(English::add_possessive("dog"), "dog's");
assert_eq!(English::add_possessive("dogs"), "dogs'");
}The public API for verb conjugation and noun/adjective declension.
- Combines optimized data generated from
extractorwith inflection logic fromenglish-core - Pure Rust, no external dependencies
- Fast Binary search over pre-sorted arrays:
O(log n)lookup. - Code generation ensures no runtime penalty.
The core engine for English inflection — pure algorithmic logic.
- Implements the core rules for conjugation/declension
- Used to classify forms as regular or irregular for the extractor
- Has no data dependency — logic-only
- Can be used stand alone for an even smaller footprint (at the cost of some accuracy)
A tool to process and refine Wiktionary data.
- Parses large English Wiktionary dumps
- Extracts all verb, noun, and adjective forms
- Uses
english-coreto filter out regular forms, preserving only irregulars - Generates sorted static arrays for use in
english
This project relies on raw data extracted from Wiktionary. Current version built with data from 8/17/2025.
- Download the raw Wiktextract JSONL dump (~20 GB) from Kaikki.org.
- Place the file somewhere accessible (e.g.
../rawwiki.jsonl). - From the
extractorfolder, run:cargo run --release ../rawwiki.jsonl - Move the generated files adj_array.rs, noun_array.rs, verb_array.rs into the /src of english
Performance benchmarks were run on my M2 Macbook.
Writing benchmarks and tests for such a project is rather difficult and requires opinionated decisions. Many words may have alternative inflections, and the data in wiktionary is not perfect. Many words might be both countable and uncountable, the tagging of words may be inconsistent. This library includes a few uncountable words in its dataset, but not all. Uncountable words require special handling anyway. Take all benchmarks with a pound of salt, write your own tests for your own usecases. Any suggestions to improve the benchmarking are highly appreciated.
Wiktionary data is often unstable and subject to weird changes. This means that the provided inflections may change unexpectedly. You can look at the diffs of *_array.rs files for a source of truth.
- Ole in the bevy discord suggested I use
phfinstead of sorted arrays, this resulted in up to 40% speedups - https://github.com/atteo/evo-inflector
- https://github.com/plurals/pluralize
-
Code: Dual licensed under MIT and Apache © 2024 gold-silver-copper
-
Data: Wiktionary content is dual-licensed under