Jina Embeddings V2 Base: An Open-Source English Text Embedding Model - Free for Sentence Similarity and Feature Extraction

Jina Embeddings V2 Base En Q5 K M GGUF

Developed by djuna

Jina Embeddings V2 Base is an efficient English text embedding model focused on sentence similarity and feature extraction tasks.

Text Embedding EnglishOpen Source License:Apache-2.0 #English Text Embedding #High-precision Semantic Retrieval #Multi-task Evaluation Optimization

Downloads 85

Release Time : 7/28/2024

Model Overview

Released under the Apache 2.0 license, this model is primarily used for generating high-quality text embeddings suitable for various natural language processing tasks.

Model Features

Efficient Text Embedding

Capable of generating high-quality text embeddings suitable for various downstream tasks.

Multi-task Support

Performs well on multiple evaluation datasets, including classification, clustering, and retrieval tasks.

Open Source License

Uses Apache 2.0 license, allowing commercial use and modification.

Model Capabilities

Text feature extraction

Sentence similarity calculation

Text classification

Information retrieval

Text clustering

Use Cases

E-commerce

Product Review Classification

Classifying Amazon product reviews

Achieved 88.54% accuracy on MTEB AmazonPolarityClassification

Finance

Bank Customer Service Query Classification

Automatically classifying bank customer service queries

Achieved 84.01% accuracy on MTEB Banking77Classification

Academic Research

Paper Clustering

Topic clustering for arXiv and biorxiv papers

Achieved 45.39 v_measure score on MTEB ArxivClusteringP2P

🚀 jina-embedding-b-en-v2

This is a model for feature extraction, sentence similarity, and other related tasks. It has been evaluated on multiple datasets and tasks, showing its performance in different scenarios.

📚 Documentation

Model Information

Property	Details
Base Model	jinaai/jina-embeddings-v2-base-en
Datasets	allenai/c4
Language	English
License	apache-2.0
Tags	sentence-transformers, feature-extraction, sentence-similarity, mteb, llama-cpp, gguf-my-repo
Inference	false

Model Results

The model named jina-embedding-b-en-v2 has the following evaluation results on various tasks:

Classification Tasks

MTEB AmazonCounterfactualClassification (en):
- Accuracy: 74.73134328358209
- AP: 37.765427081831035
- F1: 68.79367444339518
MTEB AmazonPolarityClassification:
- Accuracy: 88.544275
- AP: 84.61328675662887
- F1: 88.51879035862375
MTEB AmazonReviewsClassification (en):
- Accuracy: 45.263999999999996
- F1: 43.778759656699435
MTEB Banking77Classification:
- Accuracy: 84.01298701298701
- F1: 83.31690714969382

Retrieval Tasks

MTEB ArguAna:
- MAP@1: 21.693
- MAP@10: 35.487
- MAP@100: 36.862
- MAP@1000: 36.872
- MRR@1: 21.977
- MRR@10: 35.565999999999995
- MRR@100: 36.948
- MRR@1000: 36.958
- NDCG@1: 21.693
- NDCG@10: 44.181
- NDCG@100: 49.982
- NDCG@1000: 50.233000000000004
- Precision@1: 21.693
- Precision@10: 7.248
- Precision@100: 0.9769999999999999
- Precision@1000: 0.1
- Recall@1: 21.693
- Recall@10: 72.475
- Recall@100: 97.653
- Recall@1000: 99.57300000000001
MTEB CQADupstackAndroidRetrieval:
- Multiple sets of MAP, MRR, NDCG, Precision, and Recall metrics are provided for different configurations.

Clustering Tasks

MTEB ArxivClusteringP2P:
- V-measure: 45.39242428696777
MTEB ArxivClusteringS2S:
- V-measure: 36.675626784714
MTEB BiorxivClusteringP2P:
- V-measure: 37.050108150972086
MTEB BiorxivClusteringS2S:
- V-measure: 30.15731442819715

Reranking Task

MTEB AskUbuntuDupQuestions:
- MAP: 62.247725694904034
- MRR: 74.91359978894604

STS Task

MTEB BIOSSES:
- Cosine Similarity Pearson: 82.68003802970496
- Cosine Similarity Spearman: 81.23438110096286
- Euclidean Pearson: 81.87462986142582
- Euclidean Spearman: 81.23438110096286
- Manhattan Pearson: 81.61162566600755
- Manhattan Spearman: 81.11329400456184

📄 License

This project is licensed under the apache-2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご