search
HomeBackend DevelopmentPython TutorialPractical Approaches to Key Information Extraction (Part 2)

Real-Life Key Information Extraction (Part 2): Refining Open-Source LLM Performance

Continuing from Part 1, this guide explores using open-source Large Language Models (LLMs) – Qwen2 2B and LLaMA 3.1 – for invoice information extraction, focusing on overcoming data privacy concerns and hardware limitations (RTX 3060 with 12GB VRAM).

Why Qwen2 2B and LLaMA 3.1?

The choice of these models was driven by resource constraints. Qwen2-VL-2B-Instruct, due to its efficient size, was preferred over larger 7B models. LLaMA 3.1 (8B), accessed via Ollama, was selected for its optimized long-context understanding. Other models, such as Qwen2 in Ollama (lacking image support) and LLaVA (insufficient multilingual capabilities), were deemed unsuitable.

This two-model strategy leverages Qwen2's strength in general key information extraction and LLaMA 3.1's superior long-context handling and JSON output consistency, particularly for multilingual documents. Qwen2 initially extracts raw information, which LLaMA 3.1 then refines and structures into a standardized JSON format. PaddleOCR, as in Part 1, enhances vision capabilities for Qwen2.

A Japanese Invoice Example

A Japanese invoice was used as a test case. The initial OCR process (incorporating language detection and PaddleOCR) yielded the following recognized text:

<code>Recognized Text: 
{'ori_text': '根岸 東急ストア TEL 045-752-6131 領収証 [TOP2C!UbO J3カード」 クレヅッ 卜でのお支払なら 200円で3ボイン卜 お得なカード! 是非こ入会下さい。 2013年09月02日(月) レジNO. 0102 NOO07さ と う 001131 スダフエウ卜チーネ 23 単198 1396 003271 オインイ年 ユウ10 4238 000805 ソマ一ク スモー一クサーモン 1298 003276 タカナン ナマクリーム35 1298 001093 ヌテラ スフレクト 1398 000335 バナサ 138 000112 アボト 2つ 単158 1316 A000191 タマネキ 429 合計 2,111 (内消費税等 100 現金 10001 お預り合計 110 001 お釣り 7 890', 
'ori_language': 'ja', 
'text': 'Negishi Tokyu Store TEL 045-752-6131 Receipt [TOP2C!UbO J3 Card] If you pay with a credit card, you can get 3 points for 200 yen.A great value card!Please join us. Monday, September 2, 2013 Cashier No. 0102 NOO07 Satou 001131 Sudafue Bucine 23 Single 198 1396 003271 Oinyen Yu 10 4238 000805 Soma Iku Smo Iku Salmon 1298 003276 Takanan Nama Cream 35 1 298 001093 Nutella Sprect 1398 000335 Banasa 138 000112 Aboto 2 AA 158 1316 A000191 Eggplant 429 Total 2,111 (including consumption tax, etc. 100 Cash 10001 Total deposited 110 001 Change 7 890', 
'language': 'en',}</code>

Practical Approaches to Key Information Extraction (Part 2)

This was compared against a ChatGPT baseline, demonstrating ChatGPT's superior performance in this initial test.

Qwen2 2B Results (Standalone)

Using Qwen2 independently revealed its limitations. The JSON output was incomplete and inaccurate in several fields, highlighting its struggles with consistent JSON formatting and long-context processing.

Combined Qwen2 and LLaMA 3.1 Approach

Employing LLaMA 3.1 as a post-processor to refine Qwen2's output yielded improved, but still imperfect, results. While some key fields were accurately extracted, detailed item information remained problematic.

Future Improvements: Fine-tuning Qwen2VL

The next part will detail fine-tuning the Qwen2VL model using a dataset of receipt data, aiming to significantly improve both accuracy and processing speed for this specialized task.

Conclusion

This experiment showcases the potential and limitations of open-source LLMs for real-world key information extraction. While a combined approach offers improvements over using a single model, further refinement, particularly through model fine-tuning, is necessary to achieve optimal performance. The focus on data privacy and efficient resource utilization remains a key advantage of this open-source approach.

The above is the detailed content of Practical Approaches to Key Information Extraction (Part 2). For more information, please follow other related articles on the PHP Chinese website!

Statement
The content of this article is voluntarily contributed by netizens, and the copyright belongs to the original author. This site does not assume corresponding legal responsibility. If you find any content suspected of plagiarism or infringement, please contact admin@php.cn
How to Use Python to Find the Zipf Distribution of a Text FileHow to Use Python to Find the Zipf Distribution of a Text FileMar 05, 2025 am 09:58 AM

This tutorial demonstrates how to use Python to process the statistical concept of Zipf's law and demonstrates the efficiency of Python's reading and sorting large text files when processing the law. You may be wondering what the term Zipf distribution means. To understand this term, we first need to define Zipf's law. Don't worry, I'll try to simplify the instructions. Zipf's Law Zipf's law simply means: in a large natural language corpus, the most frequently occurring words appear about twice as frequently as the second frequent words, three times as the third frequent words, four times as the fourth frequent words, and so on. Let's look at an example. If you look at the Brown corpus in American English, you will notice that the most frequent word is "th

Image Filtering in PythonImage Filtering in PythonMar 03, 2025 am 09:44 AM

Dealing with noisy images is a common problem, especially with mobile phone or low-resolution camera photos. This tutorial explores image filtering techniques in Python using OpenCV to tackle this issue. Image Filtering: A Powerful Tool Image filter

How Do I Use Beautiful Soup to Parse HTML?How Do I Use Beautiful Soup to Parse HTML?Mar 10, 2025 pm 06:54 PM

This article explains how to use Beautiful Soup, a Python library, to parse HTML. It details common methods like find(), find_all(), select(), and get_text() for data extraction, handling of diverse HTML structures and errors, and alternatives (Sel

How to Perform Deep Learning with TensorFlow or PyTorch?How to Perform Deep Learning with TensorFlow or PyTorch?Mar 10, 2025 pm 06:52 PM

This article compares TensorFlow and PyTorch for deep learning. It details the steps involved: data preparation, model building, training, evaluation, and deployment. Key differences between the frameworks, particularly regarding computational grap

Introduction to Parallel and Concurrent Programming in PythonIntroduction to Parallel and Concurrent Programming in PythonMar 03, 2025 am 10:32 AM

Python, a favorite for data science and processing, offers a rich ecosystem for high-performance computing. However, parallel programming in Python presents unique challenges. This tutorial explores these challenges, focusing on the Global Interprete

How to Implement Your Own Data Structure in PythonHow to Implement Your Own Data Structure in PythonMar 03, 2025 am 09:28 AM

This tutorial demonstrates creating a custom pipeline data structure in Python 3, leveraging classes and operator overloading for enhanced functionality. The pipeline's flexibility lies in its ability to apply a series of functions to a data set, ge

Serialization and Deserialization of Python Objects: Part 1Serialization and Deserialization of Python Objects: Part 1Mar 08, 2025 am 09:39 AM

Serialization and deserialization of Python objects are key aspects of any non-trivial program. If you save something to a Python file, you do object serialization and deserialization if you read the configuration file, or if you respond to an HTTP request. In a sense, serialization and deserialization are the most boring things in the world. Who cares about all these formats and protocols? You want to persist or stream some Python objects and retrieve them in full at a later time. This is a great way to see the world on a conceptual level. However, on a practical level, the serialization scheme, format or protocol you choose may determine the speed, security, freedom of maintenance status, and other aspects of the program

Mathematical Modules in Python: StatisticsMathematical Modules in Python: StatisticsMar 09, 2025 am 11:40 AM

Python's statistics module provides powerful data statistical analysis capabilities to help us quickly understand the overall characteristics of data, such as biostatistics and business analysis. Instead of looking at data points one by one, just look at statistics such as mean or variance to discover trends and features in the original data that may be ignored, and compare large datasets more easily and effectively. This tutorial will explain how to calculate the mean and measure the degree of dispersion of the dataset. Unless otherwise stated, all functions in this module support the calculation of the mean() function instead of simply summing the average. Floating point numbers can also be used. import random import statistics from fracti

See all articles

Hot AI Tools

Undresser.AI Undress

Undresser.AI Undress

AI-powered app for creating realistic nude photos

AI Clothes Remover

AI Clothes Remover

Online AI tool for removing clothes from photos.

Undress AI Tool

Undress AI Tool

Undress images for free

Clothoff.io

Clothoff.io

AI clothes remover

AI Hentai Generator

AI Hentai Generator

Generate AI Hentai for free.

Hot Article

R.E.P.O. Energy Crystals Explained and What They Do (Yellow Crystal)
2 weeks agoBy尊渡假赌尊渡假赌尊渡假赌
Repo: How To Revive Teammates
1 months agoBy尊渡假赌尊渡假赌尊渡假赌
Hello Kitty Island Adventure: How To Get Giant Seeds
4 weeks agoBy尊渡假赌尊渡假赌尊渡假赌

Hot Tools

SAP NetWeaver Server Adapter for Eclipse

SAP NetWeaver Server Adapter for Eclipse

Integrate Eclipse with SAP NetWeaver application server.

EditPlus Chinese cracked version

EditPlus Chinese cracked version

Small size, syntax highlighting, does not support code prompt function

Dreamweaver Mac version

Dreamweaver Mac version

Visual web development tools

Notepad++7.3.1

Notepad++7.3.1

Easy-to-use and free code editor

VSCode Windows 64-bit Download

VSCode Windows 64-bit Download

A free and powerful IDE editor launched by Microsoft