# linkedin/detext

DeText: A Deep Neural Text Understanding Framework for Ranking and Classification Tasks

Repository: https://github.com/linkedin/detext
Canonical: https://ross.abutalabs.com/products/detext
Language: Python
License: BSD-2-Clause
License Family: permissive
Topics: ranking, deep-neural-networks, detext-framework, classification, text-embeddings, nlp
Last push: 2023-03-02T22:27:06+00:00

## Health v2 (maintenance only)
Score: 23/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 100
- inputs: {"age_days": 2421, "days_push": 1280, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1263, forks 135 (observed 2026-08-28T04:04:10.444290+00:00)

## What it is
DeText is a deep neural text understanding framework from LinkedIn for NLP ranking, classification, and language generation tasks. It uses semantic matching with configurable text encoders (CNN, BERT, LSTM) combined with wide & deep feature processing, built on TensorFlow.

## Use cases
- build a learning-to-rank model for search results
- train a semantic matching model for query understanding
- classify text into multiple categories with deep neural networks
- rank items in a recommender system using text embeddings
- fine-tune BERT for a ranking task
- combine deep text features with traditional features in a wide and deep model

## When to choose
- you need a configurable TensorFlow framework for text-based ranking or classification
- you want to combine BERT/CNN/LSTM encoders with wide & deep feature processing
- you need learning-to-rank losses with semantic matching interaction layers

## When to avoid
- you need a lightweight or non-TensorFlow NLP stack
- you want actively developed tooling with frequent releases
- you need simple keyword search without deep learning

## Facets
- artifact type: framework
- maturity: maintenance
- function: machine-learning, nlp, deep-learning, search-engine
- domain: machine-learning
- platform: python
- tags: ranking, learning-to-rank, text-classification, semantic-matching, tensorflow, bert, query-understanding, natural-language-processing, search

## Member repositories
- linkedin/detext (main) score 23

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:10.444290+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 2026-08-30T05:04:08.884553+00:00, confidence not recorded.
  - readme: https://github.com/linkedin/detext (fetched 2026-08-28T04:04:10.444290+00:00, sha 629b87bf0b51)
- Data as of 2026-08-30T08:39:29.467469+00:00.