COFO: COdeFOrces dataset for Program Classification, Recognition and Tagging

2025-03-24

AI Summary

Program understanding faces challenges in modeling fine-grained, multilingual semantic representations. Method: We introduce COFO, the first large-scale, multilingual, structured competitive programming benchmark, covering 809 algorithmic problem categories and comprising 369K C/C++/Java/Python source code submissions. Each instance includes full problem statements, standardized I/O examples, and multidimensional semantic labels (e.g., algorithm type, data structure, difficulty). Our methodology systematically integrates dynamic web crawling (Selenium + BeautifulSoup), multilingual code normalization, I/O specification parsing, and hierarchical problem-label ontology modeling. Contribution/Results: Experiments demonstrate substantial improvements over baselines in program classification, intent recognition, and cross-lingual code representation learning. COFO provides a high-quality, open-source benchmark to advance foundational research and rigorous evaluation in code intelligence.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Constraint Satisfaction and Optimization: Constraint Programming
• Computer Vision: Language and Vision

Application Category

• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

In recent years, a lot of technological advances in computer science have aided software programmers to create innovative and real-time user-friendly software. With the creation of the software and the urging interest of people to learn to write software, there is a large collection of source codes that can be found on the web, also known as Big Code, which can be used as a source of data for driving the machine learning applications tending to solve certain software engineering problems. In this paper, we present COFO, a dataset consisting of 809 classes/problems with a total of 369K source codes written in C, C++, Java, and Python programming languages, along with other metadata such as code tags, problem specification, and input-output specifications. COFO has been scraped from the openly available Codeforces website using a selenium-beautifulsoup-python based scraper. We envision that this dataset can be useful for solving machine learning-based problems like program classification/recognition, tagging, predicting program properties, and code comprehension.

Problem

• Creating dataset for program classification and recognition
• Providing source codes for machine learning applications
• Solving software engineering problems with big code

Innovation

• Scraped Codeforces data using Selenium-BeautifulSoup-Python
• Created COFO dataset with 369K multi-language source codes
• Includes metadata like tags and problem specifications