K

Koe5

Developed by nlpai-lab
KoE5 is a Korean text retrieval model fine-tuned based on intfloat/multilingual-e5-large, demonstrating outstanding performance in Korean text retrieval.
Downloads 10.63k
Release Time : 9/24/2024

Model Overview

KoE5 is a Korean text retrieval model fine-tuned from the intfloat/multilingual-e5-large model using the ko-triplet-v1.0 dataset, primarily designed for Korean and English text feature extraction and retrieval tasks.

Model Features

Korean Optimization
Specifically optimized for Korean text retrieval, excelling in Korean retrieval tasks.
Multilingual Support
Supports both Korean and English text processing.
Efficient Retrieval
Provides efficient text retrieval capabilities based on the advanced E5 architecture.
Large-scale Training Data
Trained using over 700,000 Korean query-document-hard negative sample pairs.

Model Capabilities

Text Feature Extraction
Semantic Similarity Calculation
Cross-language Retrieval
Document Retrieval

Use Cases

Information Retrieval
Open-domain Q&A
Used for passage retrieval in Korean open-domain Q&A systems.
Performs well on the Ko-StrategyQA dataset.
Legal Document Retrieval
Retrieves relevant passages from a large corpus of legal documents.
Excels on the legal-domain AutoRAGRetrieval dataset.
Semantic Analysis
Semantic Similarity Calculation
Calculates the semantic similarity between two Korean texts.
Can be used for text matching, deduplication, and similar tasks.
Featured Recommended AI Models
ยฉ 2025AIbase