# babylonhealth/fastText_multilingual

Multilingual word vectors in 78 languages

Repository: https://github.com/babylonhealth/fastText_multilingual
Canonical: https://ross.abutalabs.com/products/fasttext_multilingual
Language: Jupyter Notebook
License: BSD-3-Clause
License Family: permissive
Topics: word-vectors, machine-learning, machine-translation, natural-language-processing, nlp, distributed-representations
Last push: 2023-03-10T17:39:46+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3421, "days_push": 1272, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1201, forks 120 (observed 2026-08-28T04:03:58.210476+00:00)

## What it is
A collection of 78 alignment matrices that project Facebook's monolingual fastText word vectors into a single shared multilingual vector space, based on an ICLR 2017 paper. It includes Jupyter notebooks for using the matrices and learning your own alignments, enabling cross-lingual word translation prediction.

## Use cases
- align word vectors from different languages into one shared space
- predict translations of words across languages without a bilingual dictionary
- build multilingual NLP models with cross-lingual embeddings
- compare word similarity across languages like French and Russian
- learn custom alignment matrices for new language pairs
- zero-shot cross-lingual word translation for language pairs without training data

## When to choose
- you need cross-lingual word embeddings for the original fastText 90 languages
- you want to do zero-shot word translation between language pairs without bilingual dictionaries
- you need monolingual fastText behavior preserved while gaining multilingual alignment
- you want to reproduce or extend the ICLR 2017 bilingual word vectors research

## When to avoid
- you need actively maintained tooling or support
- you need alignment for languages beyond the original fastText 90 (e.g., the newer 204 languages)
- you need sentence-level or contextual embeddings rather than static word vectors
- you want a turnkey multilingual model rather than raw matrices plus fastText vectors

## Facets
- artifact type: dataset
- maturity: abandoned
- function: nlp, machine-learning, serialization
- domain: machine-learning
- platform: python, cross-platform
- tags: word-embeddings, word-vectors, fasttext, multilingual, cross-lingual-embeddings, translation, svd-alignment, pretrained-models, natural-language-processing, algorithms

## Member repositories
- babylonhealth/fastText_multilingual (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:58.210476+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:20:39.771694+00:00, confidence not recorded.
  - readme: https://github.com/babylonhealth/fastText_multilingual (fetched 2026-08-28T04:03:58.210476+00:00, sha d173cb110fdb)
- Data as of 2026-08-30T08:39:29.467469+00:00.