Automated Metadata Generation for Multi-lingual Content Evaluation
detail.loadingPreview
This workflow automatically generates structured metadata in both English and Chinese by evaluating LLM-generated summaries against source transcripts.
About This Workflow
This n8n workflow is designed to automate the generation of structured metadata for content evaluation, specifically focusing on generating summaries and then evaluating their quality in a multi-lingual context. It leverages various n8n nodes, including webhook triggers, Google Drive integration, text extraction, LLM summarization (using OpenAI and Google Gemini), and an evaluation node that assesses instruction following, groundedness, conciseness, and fluency. The workflow ultimately aims to produce bilingual metadata (English and Chinese) reflecting the evaluation results.
Key Features
- Multi-lingual Metadata Generation: Outputs structured metadata in both English and Chinese.
- Automated Summarization: Leverages LLMs (OpenAI, Gemini) to summarize video transcripts.
- Content Evaluation: Assesses the quality of AI-generated summaries based on predefined criteria (instruction following, groundedness, conciseness, fluency).
- Google Drive Integration: Downloads video transcripts from Google Drive.
- Google Sheets Integration: Fetches data and writes evaluation results back to Google Sheets.
- Webhook Trigger: Initiates the workflow with a Google Drive URL.
- Structured Output: Generates JSON output for evaluation metrics and scores.
How To Use
- Setup Trigger: Configure the
Webhooknode to receive agdrive_urlin the request body. - Set Gdrive URL: The
Get Gdrive URLnode (type: Set) extracts thegdrive_urlfrom the webhook payload. - Download Transcript: The
Download Transcriptnode (type: Google Drive) downloads the content of the file specified by thegdrive_url. - Extract Text: The
Extract from Filenode (type: Extract from File) extracts the text content from the downloaded file. - Summarize Content: The
Summarise Agentnode (type: Chain LLM) uses an LLM (currently configured with OpenAI Chat Model) to summarize the extracted transcript. - Evaluate Summarization: The
Evaluate Summarisationnode (type: Chain LLM) uses another LLM (currently configured with Google Gemini) to evaluate the quality of the summary against the original transcript based on provided criteria. - Set Outputs: The
Set Outputsnode (type: Evaluation) takes the evaluation results (scoreandreason) and prepares them for writing to Google Sheets. - Set Metrics: The
Set Metricsnode (type: Evaluation) assigns a numericalscorebased on the evaluation output. - Respond to User/Is Evaluating?: The
Is Evaluating?node (type: Evaluation) acts as a conditional check. If evaluations are active, it proceeds to further processing; otherwise, it might respond to the user (viaRespond to Usernode, though its connection is shown but not fully detailed for this path). - Output Parsers: The
Outputparser (type: @n8n/n8n-nodes-langchain.outputParserStructured) is used to structure the output from theLLMnode. - Dataset Row Fetching: The
When fetching a dataset rownode (type: Evaluation Trigger) is configured to read from a specific Google Sheet ('Summarization' sheet in '96. Evaluations Test' document) and likely triggers the workflow with data from that row.
Apps Used
Workflow JSON
{
"id": "905f7e41-355c-4e1c-908b-f6a3f3d9b558",
"name": "Automated Metadata Generation for Multi-lingual Content Evaluation",
"nodes": 17,
"category": "AI",
"status": "active",
"version": "1.0.0"
}Note: This is a sample preview. The full workflow JSON contains node configurations, credentials placeholders, and execution logic.
Get This Workflow
ID: 905f7e41-355c...
About the Author
N8N_Community_Pick
Curator
Hand-picked high quality workflows from the global community.
Statistics
Related Workflows
Discover more workflows you might like
Build an AI Documentation Expert Chatbot with Gemini RAG
This n8n workflow automates the creation of an AI-powered expert chatbot capable of answering questions based on your documentation. It intelligently ingests, cleans, and processes your knowledge base, preparing it for a Retrieval Augmented Generation (RAG) pipeline with Google Gemini.
Generate Multilingual Image Captions and Overlay
This workflow generates captions for images using Google Gemini and overlays them onto the image, with support for multilingual output.
LangChain Workflow Retriever Example
Demonstrates using LangChain's Retriever QA Chain to query data retrieved from another workflow.