MCINext/beytoote-clustering

Dataset Summary

Beytoote Clustering (BeytooteClustering) is a Persian (Farsi) dataset created for the Clustering task, specifically for grouping blog articles across a wide range of topics. It is part of the FaMTEB (Farsi Massive Text Embedding Benchmark). The dataset was sourced from beytoote.com, a Persian-language blog site, and includes content organized into 19 distinct thematic categories.

• Language(s): Persian (Farsi)
• Task(s): Clustering (Document Clustering, Topic Modeling)
• Source: Crawled from the Beytoote (beytoote.com) blog platform
• Part of FaMTEB: Yes

Supported Tasks and Leaderboards

This dataset is used to benchmark text embedding models on clustering performance over diverse, real-world Persian web content. It evaluates how well models can group thematically similar blog posts. Results appear on the Persian MTEB Leaderboard under the Clustering task.

Construction

1. Around 200,000 blog documents were initially scraped from beytoote.com.
2. Articles were automatically categorized using tags found in their URLs, resulting in 22 initial groups.
3. Only the content inside each post’s <summary> tag was retained to ensure consistency in length.
4. Categories with fewer than 500 samples were excluded to improve balance and reliability.
5. The final dataset includes 95,851 samples across 19 categories.

Data Splits

• Train: 0 samples
• Development (Dev): 0 samples
• Test: 95,851 samples