Chandu06/bucket / README.md
Chandu06's picture
|
download
raw
8.08 kB
---
license: apache-2.0
language:
- en
task_categories:
- text-generation
- text-retrieval
- feature-extraction
pretty_name: Resume JSON Extraction Dataset
dataset_type: synthetic
description: >
This dataset contains structured, transformed representations of resumes
created for research and model training purposes. It does not distribute
original proprietary resume documents verbatim.
tags:
- information-extraction
- resume
- json
- llm
- qwen
---
# Dataset Card for resume-json-extraction-5k
## Dataset Description
This dataset contains 4,879 resume examples formatted for fine-tuning language models to extract structured JSON information from resume text.
### Dataset Summary
The dataset consists of resume text paired with structured JSON outputs containing:
- Job titles (current and previous)
- Companies (current and previous)
- Years of experience
- Seniority level
- Primary domain and industries
- Core and secondary skills
- Tools used
- Leadership experience
- Key achievements
- Location
- Professional summary
This dataset is designed for supervised fine-tuning of language models to perform information extraction tasks. Each training example consists of:
- **Input**: Raw resume text (unstructured)
- **Output**: Structured JSON representation with extracted professional information
The dataset consists of resume text paired with structured JSON outputs containing comprehensive professional profile information extracted from resumes.
### What This Dataset Contains
Each example in this dataset represents a complete conversation formatted for training:
1. **System Message**: Instructions for the model to act as an expert resume parser
2. **User Message**: The raw resume text to be parsed
3. **Assistant Message**: The structured JSON output with extracted information
The data is formatted using Qwen3's chat template (`<|im_start|>` and `<|im_end|>` tokens) and is ready for direct use in fine-tuning pipelines.
### Languages
The dataset is in English.
## Dataset Structure
### Data Instances
Each instance contains a single field `text` with the complete chat template formatted conversation:
```json
{
"text": "<|im_start|>system\nYou are an expert resume parser. Extract structured information from resumes and return ONLY valid JSON. Do not include explanations or extra text.<|im_end|>\n<|im_start|>user\nResume:\n[Raw resume text here...]\n<|im_end|>\n<|im_start|>assistant\n{\"current_title\": \"...\", \"previous_titles\": [...], ...}<|im_end|>"
}
```
### Example Data Structure
Here's what a complete example looks like:
**Input (User Message)**:
```
Resume:
SOFTWARE ENGINEER
John Doe
Email: john@example.com
Summary:
Experienced software engineer with 5 years in Python and web development.
Experience:
Senior Software Engineer | Tech Corp | 2020 - Present
- Developed REST APIs using Python and Flask
- Led team of 3 junior developers
...
```
**Output (Assistant Message - Structured JSON)**:
```json
{
"current_title": "Senior Software Engineer",
"previous_titles": ["Software Engineer"],
"current_company": "Tech Corp",
"previous_companies": ["Previous Company"],
"years_experience": 5,
"seniority": "senior",
"primary_domain": "Software Engineering",
"industries": ["Technology", "SaaS"],
"core_skills": ["Python", "REST APIs", "Web Development"],
"secondary_skills": ["Team Leadership", "Flask", "API Design"],
"tools": ["Python", "Flask", "Git"],
"leadership_experience": true,
"key_achievements": [
"Led team of 3 junior developers",
"Developed REST APIs serving 1M+ requests/day"
],
"location": null,
"summary": "Experienced software engineer with 5 years in Python and web development; led teams and built scalable REST APIs."
}
```
### Data Fields
- **text**: Complete chat template formatted text with system prompt, user resume, and assistant JSON response
### Structured JSON Schema
The assistant output in each example contains a structured JSON object with the following fields:
#### Core Profile Fields
- **`current_title`** (string): The most recent or current job title
- **`previous_titles`** (array of strings): List of prior job titles (up to 3 most recent)
- **`current_company`** (string): The most recent or current company name
- **`previous_companies`** (array of strings): List of prior companies (up to 3 most recent)
- **`years_experience`** (number): Total years of professional experience (e.g., 7.5)
- **`seniority`** (string): Seniority level - one of: `"junior"`, `"mid"`, `"senior"`, `"lead"`, or `"principal"`
#### Domain and Industry
- **`primary_domain`** (string): Primary professional domain or field (e.g., "Software Engineering", "Marketing")
- **`industries`** (array of strings): List of industries the person has worked in (up to 3 most prominent)
#### Skills and Tools
- **`core_skills`** (array of strings): Primary/core skills (up to 3 most prominent)
- **`secondary_skills`** (array of strings): Secondary or supporting skills (up to 3 most prominent)
- **`tools`** (array of strings or null): Tools, software, or technologies used (up to 3 most prominent)
#### Experience and Achievements
- **`leadership_experience`** (boolean): Whether the resume indicates leadership, management, or mentoring experience
- **`key_achievements`** (array of strings): Impact-oriented achievements with metrics or outcomes (up to 3 highest-impact)
#### Additional Information
- **`location`** (string or null): Geographic location if mentioned in the resume
- **`summary`** (string): Concise professional summary (maximum 300 characters)
#### Field Constraints
- Array fields are limited to a maximum of 3 items (most recent or most relevant)
- Fields may be `null` if information is not present or cannot be confidently determined
- All information is extracted only from what is explicitly stated in the resume (no inference)
### Data Splits
- **train**: 4,879 examples (100% of dataset)
## Dataset Creation
### Source Data
The dataset was created from extracted resume data where:
1. Raw resume text was collected
2. Structured information was extracted using LLM-based extraction
3. Examples were formatted using Qwen3 chat template
4. Only successful extractions (no errors) were included
### Preprocessing
1. Filtered out examples with extraction errors
2. Converted to Qwen3 chat template format
3. Applied tokenizer chat template with:
- `add_generation_prompt=False` (for training)
- `enable_thinking=False` (for JSON output)
### Annotations
Each example includes:
- System prompt: Instructions for resume parsing
- User input: Raw resume text
- Assistant output: Structured JSON with extracted information
## Uses
### Direct Use
This dataset is intended for fine-tuning language models to:
- Extract structured information from resumes
- Generate JSON output from unstructured text
- Perform information extraction tasks
### Out-of-Scope Use
This dataset should not be used for:
- Training models to generate fake resumes
- Extracting personal information for unauthorized purposes
- Any use case that violates privacy or data protection laws
## Bias, Risks, and Limitations
### Bias
- Dataset may reflect biases present in the source resume data
- Geographic and industry representation may be uneven
- May not represent all resume formats or styles
### Risks
- Models trained on this data may perpetuate biases
- Extracted information should be verified for accuracy
- Not suitable for automated hiring decisions without human review
### Limitations
- Limited to ~5k examples
- May not generalize to all resume formats
- JSON structure is fixed and may not capture all resume information
## Citation
```bibtex
@misc{resume_json_extraction_5k,
title={resume-json-extraction-5k},
author={Sandeep Panem},
year={2026},
publisher={Hugging Face},
howpublished={\url{https://huggingface.co/datasets/sandeeppanem/resume-json-extraction-5k}}
}
```
## License
This dataset is licensed under Apache 2.0.
## Contact
For questions or issues, please open an issue on the dataset repository.

Xet Storage Details

Size:
8.08 kB
·
Xet hash:
ac83b9906b2c9c39b46563841c5c800ba77ceab1eac1094f0a1bc9f2917ff7d0

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.