Hugging Face Paper Analysis and Notion Archiving

Overview

Extracting and organizing daily research papers from Hugging Face is tedious and error-prone. This workflow automates scraping paper lists by date, extracts key details using a large language model, and archives structured entries in Notion—eliminating manual data entry and boosting retrieval efficiency.

The Impact

Skip manual scraping. Automatically fetch and parse daily Hugging Face paper lists without user intervention.
Flag duplicates. Cross-check Notion entries by URL to avoid redundant archiving.
Extract insights. Use LLM to summarize abstracts into structured JSON with keywords, metrics, and classifications.
Archive instantly. Create fully formatted Notion pages with all key paper metadata for easy future access.

Who This Is For

Researchers automating daily paper tracking and note organization in Notion.
R&D Teams building centralized literature libraries with abstracts and keywords.
Product Managers monitoring tech trends by capturing emerging methods and performance data.
Data Scientists wanting structured access to up-to-date research insights.

How It Works

Calculate Query Dates
Compute yesterday’s date for targeted paper scraping and today’s date for scrap record.

Scrape Paper URLs
Send GET request to Hugging Face papers page filtered by date; extract unique paper URLs from HTML anchors with class "line-clamp-3".

Deduplicate and Check Notion
Loop over URLs; query Notion database by URL to skip existing entries and process only new papers.

Fetch Paper Details
Request full paper HTML; extract title (class "text-2xl") and abstract (class "text-gray-600" or fallback selectors) via regex parsing.

Analyze and Archive
Use LLM to generate structured JSON with core introduction, keywords, results, technical details, classification; format and write these into Notion with proper field types.

What You'll Need

Before using this template, make sure you have:

A Notion database preconfigured with exact attribute names: title (Title), URL (URL), abstract (Rich Text), scrapdate (Date), Classification (Rich Text), TechnicalDetails (Rich Text), DataAndResults (Rich Text), keywords (Rich Text), CoreIntroduction (Rich Text).
Valid Notion API credentials authorized to create and query pages in your target database.
The NotionDatabaseId extracted from your database URL to link the workflow correctly.

How to Use

Step 1. Prepare Notion Database

Create a database with all required fields matching attribute names exactly as specified.

Step 2. Configure Credentials

Set up Notion API credentials in the workflow’s tool for authorized access.

Step 3. Input Database ID

Enter your NotionDatabaseId parameter with the target database’s ID from its URL.

Step 4. Run Workflow

Trigger the workflow. It will scrape yesterday’s Hugging Face papers, analyze abstracts, and create Notion entries.

Step 5. Verify Results

Check your Notion database to confirm new paper entries are created with all structured metadata.

FAQs

How does the workflow avoid duplicating existing paper entries in Notion?

It queries the Notion database for each paper URL before processing. If an entry with the same URL exists, it skips scraping and writing that paper again.

Which HTML elements does the workflow target to extract paper titles and abstracts?

It extracts titles from <h1> tags with class "text-2xl" and abstracts primarily from <p> tags with class "text-gray-600". It has fallback regex patterns for alternate abstract locations.

How does the LLM contribute to this workflow?

The large language model analyzes the extracted abstract text, producing structured JSON with core introduction, keywords, key results, technical details, and classification for rich metadata archiving.

What must be configured in Notion to use this workflow?

You need a Notion database with fields named exactly as required (title, URL, abstract, scrapdate, Classification, TechnicalDetails, DataAndResults, keywords, CoreIntroduction) and valid Notion API credentials to access it.

Updated on: Aug 21, 2026

Was This Page Helpful?