README.md

---
license: cc-by-4.0
task_categories:
- text-retrieval
language:
- en
configs:
- config_name: Papers
  data_files:
  - split: papers_collection
    path:
    - papers_collection.jsonl
  - split: papers_test
    path:
    - papers_test.jsonl
  - split: papers_test_judgeable
    path:
    - papers_test_judgeable.jsonl
- config_name: Qrels
  data_files:
  - split: qrels_authors
    path:
    - qrels/qrels.test.authors.tsv
  - split: qrels_cite
    path:
    - qrels/qrels.test.cite.tsv
  - split: qrels_simcite
    path:
    - qrels/qrels.test.simcite.tsv
  sep: "\t"
tags:
- reviewer-assignment
- scientific-papers
- authorship
- information-retrieval
pretty_name: exHarmony
size_categories:
- 1M<n<10M
---


# exHarmony: Authorship and Citations for Benchmarking the Reviewer Assignment Problem

Quick links: 📃 [Paper](https://arxiv.org/pdf/2502.07683) | ⚙️ [Code](https://github.com/Reviewerly-Inc/exHarmony) 

## Dataset Summary

**exHarmony** is a large-scale benchmark dataset for the **Reviewer Assignment Problem (RAP)**, reframing reviewer recommendation as an **information retrieval** task.
It leverages publication metadata from **OpenAlex** to construct a collection of papers, their authors, citation links, and multiple **qrel definitions** for evaluation.

The dataset allows researchers to systematically study reviewer recommendation under different assumptions of reviewer expertise (e.g., authorship, citation networks, and similarity-filtered citations).

exHarmony was introduced in the paper:

> *exHarmony: Authorship and Citations for Benchmarking the Reviewer Assignment Problem*
> Ebrahimi, Salamat, Arabzadeh, Bashari, Bagheri (ECIR 2025)

* **Collection split**: A large set of scientific papers used for indexing.
* **Test split**: Papers held out for evaluation.
* **Authors' works mapping**: Links each author to their published works.
* **Authors' information**: Includes metadata such as citation counts, institutional affiliation, and years of experience.


## Usage

```python
from datasets import load_dataset

dataset = load_dataset("Reviewerly/exHarmony", "Papers") # Select data type from ['Papers', 'Qrels']

# Example: Access paper collection
papers = dataset["papers_collection"]
print(papers[0])

# Example: Access paper collection
qrels = dataset["qrels_authors"]
print(qrels[0])
```

## Dataset Structure

### Data Files

| Description            | File Name                            | File Size | Num Records | Format                                                             |
| ---------------------- | ------------------------------------ | --------- | ----------- | ------------------------------------------------------------------ |
| Collection             | `papers_collection.jsonl`            | 1.6 GB    | 1,204,150   | paper\_id, title, abstract                                         |
| Test                   | `papers_test.jsonl`                  | 15 MB     | 9,771       | paper\_id, title, abstract                                         |
| Test (judgable)        | `papers_test_judgable.jsonl`         | 14 MB     | 7,944       | paper\_id, title, abstract                                         |
| Authors’ Works Mapping | `authors_works_collection_ids.jsonl` | 222 MB    | 1,589,723   | author\_id, list\_of\_authors\_papers                              |
| Authors’ Information   | `authors_info.jsonl`                 | 225 MB    | 1,589,723   | author\_id, citation, works\_count, experience\_years, institution |

**Format:** JSON Lines (`.jsonl`), one JSON object per record.

### Example Records

**Paper record:**

```json
{"id": "https://openalex.org/W4323317762", "title": "Sharding-Based Proof-of-Stake Blockchain Protocols: Key Components & Probabilistic Security Analysis", "abstract": "Blockchain technology has been gaining great interest from a variety of sectors including healthcare, supply chain, and cryptocurrencies..."}
```

**Author works mapping:**

```json
{"id": "https://openalex.org/A5083262615", "works": ["https://openalex.org/W4323317762", "https://openalex.org/W4285189682"]}
```

**Author information:**

```json
{"id": "https://openalex.org/A5083262615", "citations": 238, "works_count": 14, "experience_years": 5, "institution": "Université de Montréal"}
```


## Qrel Files

exHarmony provides **multiple qrels** to evaluate RAP under different assumptions:

| Qrel Set              | Description                                                                   |
| --------------------- | ----------------------------------------------------------------------------- |
| **exHarmony-Authors** | Authors of each paper are considered relevant.                                |
| **exHarmony-Cite**    | Authors of the most similar paper in the test set are considered relevant.    |
| **exHarmony-SimCite** | Narrows citation-based relevance to the top-10 most similar cited papers.     |


## Citation

If you use this resource, please cite our paper:

```
@inproceedings{ebrahimi2025exharmony,
  author = {Ebrahimi, Sajad and Salamat, Sara and Arabzadeh, Negar and Bashari, Mahdi and Bagheri, Ebrahim},
  title = {exHarmony: Authorship and Citations for Benchmarking the Reviewer Assignment Problem},
  year = {2025},
  isbn = {978-3-031-88713-0},
  publisher = {Springer-Verlag},
  doi = {10.1007/978-3-031-88714-7_1},
  booktitle = {Advances in Information Retrieval: 47th European Conference on Information Retrieval, ECIR 2025, Lucca, Italy, April 6–10, 2025, Proceedings, Part III},
  pages = {1–16},
}
```