qinyan-paper-analysis

Qinyan Academic - Paper Analysis

Overview

This skill conducts in-depth analysis and interpretation of a single academic paper through the paper analysis interface of the Qinyan Academic OpenAPI. The system obtains the paper content (via PDF, webpage, or abstract) and generates a structured research analysis report, including research objectives, methodology, main findings, and research limitations.

Academic Integrity Statement: The analysis results are generated based on the actual content of the paper and will not fabricate or guess information not contained in the paper. If the complete paper content cannot be obtained, a limited analysis will be conducted based on the available information (such as the abstract), and the user will be clearly informed of the limitations of the analysis.

Prerequisites

Before use, please ensure:

1. Already set QINYAN_API_KEY environment variable (go to https://platform.qinyanai.com/ to apply)
2. Available in the environment curl and python3

export QINYAN_API_KEY="your-api-key-here"

API Interface

Interface address: https://api.qinyanai.com/v1/paper-search/analyze

Request Parameters

Parameter	Type	Required	Description
title	string	Yes	Paper title (1-1024 characters)
authors	string[]	Yes	Author list
abstract	string	No	Paper abstract (maximum 10000 characters)
doi	string	No	DOI identifier
source_url	string	No	Paper page URL
pdf_url	string	No	PDF direct link (highest priority)
language	string	No (default "Chinese")	Output language: "Chinese" or "en"

Note: abstract、source_url、pdf_url At least one of these must be provided so that the system can obtain the paper content.pdf_url has the highest priority and can obtain the most complete analysis results.

Usage

Call via scripts/analyze.sh script:

bash scripts/analyze.sh '<json_payload>'

Examples

Basic analysis (via DOI):

bash scripts/analyze.sh '{"title": "Attention Is All You Need", "authors": ["Ashish Vaswani", "Noam Shazeer", "Niki Parmar"], "doi": "10.48550/arXiv.1706.03762", "language": "中文"}'

Analysis via PDF link:

bash scripts/analyze.sh '{"title": "BERT: Pre-training of Deep Bidirectional Transformers", "authors": ["Jacob Devlin"], "pdf_url": "https://arxiv.org/pdf/1810.04805.pdf", "language": "中文"}'

Analysis via abstract:

bash scripts/analyze.sh '{"title": "论文标题", "authors": ["作者1", "作者2"], "abstract": "本文提出了...", "language": "中文"}'

English output:

bash scripts/analyze.sh '{"title": "Paper Title", "authors": ["Author1"], "doi": "10.xxxx/xxxxx", "language": "en"}'

Return Results

The API returns structured analysis results:

{
  "success": true,
  "message": "Paper analysis completed successfully",
  "paper_title": "论文标题",
  "analysis": {
    "研究目标": "提出一种新的...",
    "方法论": "设计了...",
    "主要发现": "实验结果表明...",
    "研究限制": "本研究主要在...方面存在局限"
  }
}

Output Format

When presenting the analysis results to the user, organize them according to the following structure:

1. Basic Paper Information

   • Title, authors, DOI, etc.

2. Research Objectives

   • The core problem the paper aims to solve

3. Methodology

   • Research methods used, technical approach, experimental design

4. Main Findings

   • Core research results and conclusions

5. Research Limitations

   • Limitations and shortcomings of the research

Workflow Suggestions

1. If the user only provides the paper title, first try to obtain complete information such as the DOI and abstract through literature search
2. Prioritize pdf_url(most complete analysis), then source_url, and finally abstract
3. After the analysis is completed, you may suggest to the user whether further literature search or related paper recommendations are needed