wnram — An In-Memory Go Library for WordNet

A Go library for accessing [Open English WordNet][https://en-word.net] (and Princeton WordNet compatible datasets).

Implementation Overview

This library is a native Go parser for WordNet that stores the entire database in RAM. This approach gives faster lookup times; the dataset sits at roughly 80–90 MB of RAM, which is acceptable for most server environments. Parsing the full data files takes around two seconds on a modest laptop.

Supported Features

• Lookup by term, with automatic morphological normalization (plurals, verb forms, comparatives)
• Synonyms
• All relation types (Antonym, Hyponym, Hypernym, Attribute, Entailment, etc.)
• Iteration over the full database, optionally filtered by part of speech
• Lemmatization — find the canonical form of a word
• Morphology — derive a lemma from inflected input text

Example Usage

import (
    "log"

    "github.com/coreruleset/wnram"
)

func main() {
    wn, err := wnram.New("./path/to/wordnet/data")
    if err != nil {
        log.Fatalf("failed to load wordnet: %s", err)
    }

    // Look up "yummy" restricted to adjectives
    found, err := wn.Lookup(wnram.Criteria{
        Matching: "yummy",
        POS:      wnram.PartOfSpeechList{wnram.Adjective},
    })
    if err != nil {
        log.Fatalf("%s", err)
    }

    // Dump details about each matching synset to console
    for _, f := range found {
        f.Dump()
    }
}

Iterating the database

err := wn.Iterate(wnram.PartOfSpeechList{wnram.Noun}, func(l wnram.Lookup) error {
    fmt.Println(l.Word(), "—", l.Gloss())
    return nil
})

Querying relations

found, _ := wn.Lookup(wnram.Criteria{Matching: "good", POS: wnram.PartOfSpeechList{wnram.Adjective}})
for _, f := range found {
    for _, a := range f.Related(wnram.Antonym) {
        fmt.Println("antonym:", a.Word())
    }
}

Developer Guide

Prerequisites

• Go 1.26 or later

Getting the WordNet data files

The library ships with [Open English WordNet][https://github.com/globalwordnet/english-wordnet] data in the data/ folder. If you need to update or replace the dataset, place the standard WordNet data.*, *.exc, and index files in data/ and point wnram.New() at it.

Running the tests

go test ./...

With coverage:

go test -coverprofile=coverage.out ./...
go tool cover -html=coverage.out   # open coverage report in browser

Linting

The project uses golangci-lint:

golangci-lint run