LongEval 2025

This repository instruction about the Collection (French original Language) for the LongEval 2025 Information Retrieval Challenge.
The dataset supports training and evaluation in Information Retrieval (IR) tasks.

🔗 Official Challenge Website: LongEval 2025

---

📂 Folder Structure

The dataset is structured as follows:

release_2025/
│-- French/
│   │-- LongEval Train Collection/   # Training data (first release)
│   │   │-- Json/                    # JSON-formatted document collection
│   │   │-- qrels/                   # Relevance judgments (Qrels), divided by month
│   │   │   │-- 2022-06_fr/          # Qrels for June 2022
│   │   │   │   ├── qrels_processed.txt  # Processed Qrels file
│   │   │   │-- 2022-07_fr/          # Qrels for July 2022
│   │   │   │-- ...                  # Other months
│   │   │-- Trec/                    # TREC-formatted document collection, divided by month
│   │   │   │-- 2022-06_fr/          # Documents for June 2022
│   │   │   │-- 2022-07_fr/          # Documents for July 2022
│   │   │   │-- ...                  # Other months
│   │-- LongEval Test Collection/    # Test data (to be released later)
│   │-- collection_db.db             # Database mapping document IDs to URLs (Train + Test data)
│   │-- queries_db.db                # Database mapping query IDs to query texts
│   │-- queries.txt                  # Plain text file with queries and IDs
│   │-- inspect_db.py                # Script to inspect the document database
│   │-- statistics_collection.py     # Script to generate dataset statistics

---

📖 Dataset Description

1️⃣ LongEval Train Collection

• Json/: Document collection in JSON format.
• Trec/: Document collection in TREC format.
• qrels/: Relevance judgments mapping queries to relevant documents, organized by month.

2️⃣ LongEval Test Collection

• The test data will be released at a later stage.

3️⃣ Database Files

• collection_db.db: SQLite database mapping document IDs to their URLs (Train + Test data).
• queries_db.db: SQLite database linking query IDs to query texts.
• queries.txt: A plain text file listing all queries.

---

🛠 How to Use inspect_db.py

The inspect_db.py script allows participants to inspect the document collection database (collection_db.db).
It can retrieve information such as:

• The URL of a document.
• The last update timestamp of a document.
• The month(s) in which the document appears.

▶️ Usage Example

To check the details of a document by its ID, run:

python inspect_db.py collection_db.db --id <DOCUMENT_ID>

Example output:

Tables in the database: [('mapping',), ('sqlite_sequence',)]

Structure of table 'mapping':
(0, 'id', 'INTEGER', 0, None, 1)
(1, 'url', 'TEXT', 0, None, 0)
(2, 'last_updated_at', 'TEXT', 0, None, 0)
(3, 'date', 'TEXT', 0, None, 0)

Record with ID 1:
{
    "id": 1,
    "url": "https://www.blogduvoyage.fr/roadtrip-usa-conseils/",
    "last_updated_at": [1640160479, 1640160479],
    "date": ["2022-06", "2022-06"]
}

---

📌 About the LongEval Challenge

This dataset is part of the LongEval 2025 challenge, designed to evaluate longitudinal information retrieval over time.
Participants will work with time-based IR tasks, analyzing evolving document collections and queries.

📢 More details: LongEval 2025 Challenge

---