Funcom-Java-Long: 170k Java Methods with Summaries

funcom-java-long is a dataset for code summarization presented by Bansal et al. It includes 170,000 Java methods and summaries for training and 8,192 Java methods in a test set. The dataset was extracted from jm52m and has been cleaned and deduplicated.

Use Cases

• Fine-tune language models for code summarization based on the provided Java method and comment pairs.
• Train models to generate method-level documentation based on the example prompt structure.
• Benchmark code-to-text generation models using the provided training and test splits.

Strengths

• Contains 170,000 Java method and summary pairs for training.
• Includes a separate test set of 8,192 Java methods.
• The data has been cleaned and deduplicated from its source.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last updated 2023-05-11 20:25:43; freshness should be verified.
• Row count for the training set beyond the 170k figure is unknown, which may limit suitability assessment.

Provenance

Source

Extracted from jm52m dataset.

Collection Method

Cleaned and deduplicated.

Freshness

Last updated 2023-05-11 20:25:43.