iPad Shortcuts × OpenAI API × Notion Automated Vocabulary Registration Workflow -- 2
Budget: $10 – $30 USD
Project Overview
I would like to build a workflow that lets me select English words or sentences from a PDF (or manually input text), send them to the OpenAI API for linguistic analysis, parse the resulting JSON, and automatically register the data into a Notion database.
The workflow must run inside the iPad/iOS Shortcuts app and support both PDF selection and manual input.
Required Workflows
1. Shortcut A — From PDF sharing (Main flow)
User selects text in a PDF viewer
(Notion PDF, GoodNotes, Apple Books, PDF Expert, etc.)
Launches the shortcut from the Share Sheet
Shortcut sends text to OpenAI API
Receives structured JSON
Parses JSON safely
Sends processed data to Notion DB
Creates a new row
2. Shortcut B — Manual Input
Prompts user to input a word or sentence manually
Reuses the same processing logic as Shortcut A
3. Shortcut C — Core Processor (Subroutine)
Handles:
OpenAI API request
JSON parsing
CEFR classification
Sentence splitting
Duplicate checking
Notion database registration
Reused by the other two shortcuts
Functional Requirements
① CEFR Difficulty Level Classification
Determine CEFR level (A1–C2) for the word or phrase.
Save the CEFR value into a Notion Select field.
CEFR options must be fixed:
A1 / A2 / B1 / B2 / C1 / C2
② Multi-Sentence Handling
If the selected text contains multiple sentences:
Normalize and clean the text
Split safely into individual sentences
Ensure example generation and analysis run correctly per sentence
Final JSON should remain stable and parse-safe
③ Duplicate Record Check
Before creating a new Notion row:
Search the database for the same Word value
If found → Display
“This word already exists. Do you want to add it anyway?”
User chooses Register or Cancel
④ JSON Output Requirements
The OpenAI API must return JSON only, no extra text.
Required JSON keys:
{
"word": "",
"meaning_ja": "",
"pronunciation": "",
"example_en": "",
"example_ja": "",
"part_of_speech": "",
"cefr": ""
}
⑤ Notion Database Mapping
The Notion database contains these properties (in this order):
Word — Title
Meaning — Rich text
Pronunciation — Rich text
Example (EN) — Rich text
Example (JP) — Rich text
Part of speech — Select
CEFR — Select
Important:
In the future, I may add or rearrange Notion properties.
The shortcuts must continue working even if property order changes (property names will remain the same).
Technical Requirements
OpenAI API
Endpoint: POST https://api.openai.com/v1/chat/completions
Model: gpt-4.1-mini (or recommended)
Strict JSON output enforcement
Notion API
Endpoint: POST https://api.notion.com/v1/pages
Version: 2022-06-28 or latest stable
Must support property mapping and stable error handling
Deliverables
Shortcut A: “Add from PDF to Notion”
Shortcut B: “Add Word to Notion (Manual Input)”
Shortcut C: “Core Vocabulary Processor”
Short demo video (screen recording)
Brief usage manual (1–2 pages):
How to use the shortcuts
How to replace API keys
How to adjust Notion DB settings
I would like to build a workflow that lets me select English words or sentences from a PDF (or manually input text), send them to the OpenAI API for linguistic analysis, parse the resulting JSON, and automatically register the data into a Notion database.
The workflow must run inside the iPad/iOS Shortcuts app and support both PDF selection and manual input.
Required Workflows
1. Shortcut A — From PDF sharing (Main flow)
User selects text in a PDF viewer
(Notion PDF, GoodNotes, Apple Books, PDF Expert, etc.)
Launches the shortcut from the Share Sheet
Shortcut sends text to OpenAI API
Receives structured JSON
Parses JSON safely
Sends processed data to Notion DB
Creates a new row
2. Shortcut B — Manual Input
Prompts user to input a word or sentence manually
Reuses the same processing logic as Shortcut A
3. Shortcut C — Core Processor (Subroutine)
Handles:
OpenAI API request
JSON parsing
CEFR classification
Sentence splitting
Duplicate checking
Notion database registration
Reused by the other two shortcuts
Functional Requirements
① CEFR Difficulty Level Classification
Determine CEFR level (A1–C2) for the word or phrase.
Save the CEFR value into a Notion Select field.
CEFR options must be fixed:
A1 / A2 / B1 / B2 / C1 / C2
② Multi-Sentence Handling
If the selected text contains multiple sentences:
Normalize and clean the text
Split safely into individual sentences
Ensure example generation and analysis run correctly per sentence
Final JSON should remain stable and parse-safe
③ Duplicate Record Check
Before creating a new Notion row:
Search the database for the same Word value
If found → Display
“This word already exists. Do you want to add it anyway?”
User chooses Register or Cancel
④ JSON Output Requirements
The OpenAI API must return JSON only, no extra text.
Required JSON keys:
{
"word": "",
"meaning_ja": "",
"pronunciation": "",
"example_en": "",
"example_ja": "",
"part_of_speech": "",
"cefr": ""
}
⑤ Notion Database Mapping
The Notion database contains these properties (in this order):
Word — Title
Meaning — Rich text
Pronunciation — Rich text
Example (EN) — Rich text
Example (JP) — Rich text
Part of speech — Select
CEFR — Select
Important:
In the future, I may add or rearrange Notion properties.
The shortcuts must continue working even if property order changes (property names will remain the same).
Technical Requirements
OpenAI API
Endpoint: POST https://api.openai.com/v1/chat/completions
Model: gpt-4.1-mini (or recommended)
Strict JSON output enforcement
Notion API
Endpoint: POST https://api.notion.com/v1/pages
Version: 2022-06-28 or latest stable
Must support property mapping and stable error handling
Deliverables
Shortcut A: “Add from PDF to Notion”
Shortcut B: “Add Word to Notion (Manual Input)”
Shortcut C: “Core Vocabulary Processor”
Short demo video (screen recording)
Brief usage manual (1–2 pages):
How to use the shortcuts
How to replace API keys
How to adjust Notion DB settings