File size: 3,415 Bytes
05d85e4
 
 
 
 
 
 
 
 
084c2e6
fbf7b14
 
 
 
 
 
 
 
 
 
 
 
 
 
05d85e4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c33c3ac
05d85e4
 
c33c3ac
 
 
 
 
fbf7b14
 
 
c33c3ac
05d85e4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
055da42
05d85e4
a31aeef
 
 
05d85e4
 
 
 
 
 
 
 
 
 
 
084c2e6
05d85e4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
---
task_categories:
- text-retrieval
- text-ranking
language:
- en
tags:
- search
- reranking
license: cc-by-4.0
configs:
- config_name: annotated_queries
  data_files:
  - split: train
    path: annotated_queries.parquet
  default: true
- config_name: knowledge_base
  data_files:
  - split: corpus
    path: knowledge_base.parquet
- config_name: test_queries
  data_files:
  - split: test
    path: test_queries.parquet
---

# DevRev Search Dataset

## Summary

The **DevRev Search Dataset** contains transformed and anonymized user queries on DevRev's public articles, paired with relevant article chunks that answer them. It is designed to benchmark enterprise search and reranking systems. The dataset also includes test queries along with the underlying article chunks used for retrieval.

## Dataset Details

- **Curated by:** Research@DevRev  
- **Language:** English
- **Intended Use:** Evaluation of search, retrieval, and reranking models in enterprise support contexts  

## Dataset Structure

- **`annotated_queries.parquet`** — Queries paired with annotated (golden) article chunks  
- **`knowledge_base.parquet`** — Article chunks created from DevRev's customer-facing support documentation  
- **`test_queries.parquet`** — Held-out queries used for evaluation  

## Quick Start

```python
from datasets import load_dataset

annotated_queries = load_dataset("devrev/search", "annotated_queries", split="train")
knowledge_base = load_dataset("devrev/search", "knowledge_base", split="corpus")
test_queries = load_dataset("devrev/search", "test_queries", split="test")
```

## Dataset Creation

### Curation Rationale
There is a lack of publicly available datasets representing enterprise support search. This dataset addresses that gap, enabling fair and consistent evaluation of enterprise-focused search and reranking systems.

### Source Data

- **Queries:** Transformed and anonymized versions of real user queries on DevRev's public articles.  
- **Knowledge Base Articles:** Segmented (chunked) excerpts from DevRev’s customer-facing support documentation.

## Annotations

- **Annotation Process:** An automated proprietary pipeline identifies the most relevant article chunks for each query.  
- **Annotators:** Fully automated system; no human annotators were used.  
- **Validation:** Internal validation was conducted by selecting a representative sample of the automatically generated annotations, which were then manually reviewed by human evaluators to verify their correctness and ensure they meet the desired quality standards.

## Personal and Sensitive Information
All user queries were anonymized and stripped of any identifying or sensitive data. No personal information is included.

## Uses

- **Primary Use:** Training and evaluation of enterprise search, retrieval, and reranking models.  
- **Example Tasks:**  
  - [Query-to-document retrieval](https://huggingface.co/tasks/text-ranking)  
  - Reranking

## Product Impact
This dataset played a key role in honing Computer by DevRev. Learn more about Computer here: [Computer by DevRev | Your New Conversational AI Teammate](https://devrev.ai/meet-computer)

## Citation

If you use this dataset, please cite:

**BibTeX:**
```bibtex
@dataset{devrev_search_2025,
  title={DevRev Search Dataset},
  author={Research@DevRev},
  year={2025},
  url={https://huggingface.co/datasets/devrev/search},
  license={CC-BY-4.0}
}