![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F082491286038d2427f3e3f338f876e5b%2FScreenshot%202024-04-27%20at%205.04.23AM.png?generation=1714174499947051&alt=media\")
\n\n64 scalar vector => (Version 1) Machine-learned vector embedding based on document contents and metadata, where two documents that have similar technical content have a high dot product score of their embedding vectors.
\n
\n\nThe patent embeddings were built using a machine learning model that predicted a patent's CPC code from its text. Therefore, the learned embeddings are a vector of 64 continuous numbers intended to encode the information in a patent's text. Distances between the embeddings can then be calculated and used as a measure of similarity between two patents.
\n
\n\nSo, we are reverse engineering the vector to search query in this competition?
\n
\nNotebook - BigQuery for Patents
", "messages": [ { "id": "2773996", "postDate": "04/25/2024 02:46:14", "content": "
\n\n64 scalar vector => (Version 1) Machine-learned vector embedding based on document contents and metadata, where two documents that have similar technical content have a high dot product score of their embedding vectors.
\n
\n\nThe patent embeddings were built using a machine learning model that predicted a patent's CPC code from its text. Therefore, the learned embeddings are a vector of 64 continuous numbers intended to encode the information in a patent's text. Distances between the embeddings can then be calculated and used as a measure of similarity between two patents.
\n
\n\nSo, we are reverse engineering the vector to search query in this competition?
\n
\nNotebook - BigQuery for Patents
", "rawMarkdown": "\n> **64 scalar vector => (Version 1) Machine-learned vector embedding based on document contents and metadata, where two documents that have similar technical content have a high dot product score of their embedding vectors.**\n\n---\n\n[Source](https://cloud.google.com/blog/products/data-analytics/expanding-your-patent-set-with-ml-and-bigquery)\n> The **patent embeddings were built using a machine learning model that predicted a patent's CPC code from its text**. Therefore, the **learned embeddings are a vector of 64** continuous numbers intended to encode the information in a patent's text. Distances between the embeddings can then be calculated and used as a measure of similarity between two patents.\n\n---\n\n> **So, we are reverse engineering the vector to search query in this competition?**\n\n---\n\n```yaml\ntest.csv A subset of nearest_neighbors.csv that will cover 2,500 patents in the hidden dataset.\n```\n\n\n[**Notebook - BigQuery for Patents**](https://www.kaggle.com/code/jayyonamine/using-bigquery-for-patent-analysis)", "votes": null }, { "id": "2774859", "postDate": "04/25/2024 10:44:39", "content": "Yes, the goal is to explain the similarity results as structured search terms, not to build a better similarity model.
", "rawMarkdown": "Yes, the goal is to explain the similarity results as structured search terms, not to build a better similarity model.", "votes": null } ], "comments": [ { "id": 2774859, "author_name": "suicaokhoailang", "author_url": "", "post_date": "04/25/2024 10:44:39", "content": "Yes, the goal is to explain the similarity results as structured search terms, not to build a better similarity model.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2773996": "\n> **64 scalar vector => (Version 1) Machine-learned vector embedding based on document contents and metadata, where two documents that have similar technical content have a high dot product score of their embedding vectors.**\n\n---\n\n[Source](https://cloud.google.com/blog/products/data-analytics/expanding-your-patent-set-with-ml-and-bigquery)\n> The **patent embeddings were built using a machine learning model that predicted a patent's CPC code from its text**. Therefore, the **learned embeddings are a vector of 64** continuous numbers intended to encode the information in a patent's text. Distances between the embeddings can then be calculated and used as a measure of similarity between two patents.\n\n---\n\n> **So, we are reverse engineering the vector to search query in this competition?**\n\n---\n\n```yaml\ntest.csv A subset of nearest_neighbors.csv that will cover 2,500 patents in the hidden dataset.\n```\n\n\n[**Notebook - BigQuery for Patents**](https://www.kaggle.com/code/jayyonamine/using-bigquery-for-patent-analysis)", "2774859": "Yes, the goal is to explain the similarity results as structured search terms, not to build a better similarity model." }, "source": "meta" }