---
license: apache-2.0
language:
- en
---

**MedSum: A Traceable Aspect-based Summaries Benchmark for Summarizing Medical Documents**
==============
**PMID**: Unique identifier number used in PubMed.

**abstract**: Original abstracts of articles downloaded from PubMed.

**abstract_lenght**: The tokens of the original abstract.

**abstract_positioning**: Split the original abstracts into sentence lists, and attach a unique marker (position) to the header of each sentence. 

**abstract_pos_len**: The number of sentences in the abstract.

**questions_gpt**: The original answer of GPT for aspect Q: Research Questions.

**questions_summary**: The summary for aspect Q: Research Questions.

**questions_markers**: The markers for aspect Q: Research Questions.

**questions_tokens**: The number of tokens [output, input, total] using the GPT.

....