stella_en_400M_v5-FinanceRAG-v2 Open-Source Financial Model - Achieving Semantic Retrieval and Paragraph Matching for Financial Documents

Stella En 400M V5 FinanceRAG V2

Developed by thomaskim1130

A finance-domain retrieval-augmented generation model optimized based on the stella_en_400M_v5 architecture, supporting semantic retrieval and paragraph matching for financial documents

Large Language Model

Safetensors

Other#Financial Text Retrieval #Negative Sample Ranking Optimization #Domain-Specific RAG

Downloads 555

Release Time : 11/29/2024

Model Overview

This model is specifically optimized for financial document retrieval tasks, capable of understanding complex financial queries and matching relevant text passages. Trained with multiple negatives ranking loss, suitable for Q&A systems and financial information retrieval scenarios.

Model Features

Finance Domain Optimization

Specifically trained for professional content such as financial statements and terminology to enhance understanding of financial documents

Efficient Passage Retrieval

Capable of precisely locating key passages related to queries from lengthy financial documents

Multiple Negatives Training

Uses Multiple Negatives Ranking Loss to improve the ability to distinguish between similar passages

Model Capabilities

Financial document semantic retrieval

Query-passage similarity calculation

Financial Q&A system support

Key information localization in long texts

Use Cases

Financial Information Retrieval

Financial Statement Query

Query relevant statement passages based on specific financial indicators

Accurately retrieves tables and explanations containing specific financial data

Regulatory Document Analysis

Locate specific policy descriptions in SEC filings or annual reports

Quickly finds key passages related to compliance

Investment Research

Company Financial Data Extraction

Retrieve financial performance data for specific quarters or years

Precisely matches financial tables and context containing query metrics

🚀 Sentence Transformer Model

This is a sentence transformer model based on thomaskim1130/stella_en_400M_v5-FinanceRAG, capable of tasks such as sentence similarity calculation and feature extraction.

✨ Features

Sentence Similarity: Effectively calculate the similarity between sentences.
Feature Extraction: Extract features from sentences for downstream tasks.

📦 Installation

No installation steps are provided in the original document, so this section is skipped.

💻 Usage Examples

Basic Usage

The widget in the original document provides usage examples:

# Example 1
Instruct: Given a web search query, retrieve relevant passages that answer the query.
Query: Title: 
Text: What was the sum of Fourth Quarter without those Fourth Quarter smaller than 0, in 2012? (in million)

# Example 2
Instruct: Given a web search query, retrieve relevant passages that answer the query.
Query: Title: 
Text: ADBE share repurchase

📚 Documentation

Model Information

Property	Details
Model Type	Sentence Transformer
Training Data	Not provided
Dataset Size	2256
Loss	MultipleNegativesRankingLoss
Base Model	thomaskim1130/stella_en_400M_v5-FinanceRAG

Widget Examples

Example 1

Source Sentence:

Instruct: Given a web search query, retrieve relevant passages that answer the query.
Query: Title: 
Text: What was the sum of Fourth Quarter without those Fourth Quarter smaller than 0, in 2012? (in million)

Candidate Sentences:

Title: 
Text: Cash Flow Hedges Citigroup hedges the variability of forecasted cash flows associated with floating-rate assets/liabilities and other forecasted transactions.
...

Example 2

Source Sentence:

Instruct: Given a web search query, retrieve relevant passages that answer the query.
Query: Title: 
Text: ADBE share repurchase

Candidate Sentences:

Title: 
Text: TELEFLEX INCORPORATED NOTES?TO CONSOLIDATED FINANCIAL STATEMENTS — (Continued) The Company issued 82,865, 93,367 and 105,239 of non-vested restricted stock units in 2017, 2016 and 2015, respectively, the majority of which provide for vesting as to all underlying shares on the third anniversary of the grant date.
...

🔧 Technical Details

No technical details are provided in the original document, so this section is skipped.

📄 License

No license information is provided in the original document, so this section is skipped.

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