README.md

---
dataset_info:
  features:
  - name: post_discussion_id
    dtype: int64
  - name: post_discussion_tags
    sequence: string
  - name: post_discussion_title
    dtype: string
  - name: post_discussion_created_at
    dtype: timestamp[ns, tz=UTC]
  - name: post_category_id
    dtype: int64
  - name: post_discussion_views
    dtype: int64
  - name: post_discussion_reply_count
    dtype: int64
  - name: post_discussion_like_count
    dtype: int64
  - name: post_discussion_participant_count
    dtype: int64
  - name: post_discussion_word_count
    dtype: float64
  - name: post_id
    dtype: int64
  - name: post_created_at
    dtype: string
  - name: post_content
    dtype: string
  - name: post_read_count
    dtype: int64
  - name: post_reply_count
    dtype: int64
  - name: post_author_id
    dtype: string
  - name: post_number
    dtype: int64
  - name: post_discussion_related_topics
    sequence: int64
  - name: accepted_answer_post
    dtype: float64
  - name: post_content_raw
    dtype: string
  - name: post_category_name
    dtype: string
  - name: post_sentiment
    dtype: string
  - name: post_sentiment_score
    dtype: float64
  - name: post_content_cluster_embedding
    sequence: float64
  - name: post_content_classification_embedding
    sequence: float64
  - name: post_content_search_document_embedding
    sequence: float64
  - name: tag1
    dtype: string
  - name: tag2
    dtype: string
  - name: tag3
    dtype: string
  - name: tag4
    dtype: string
  - name: post_discussion_url
    dtype: string
  - name: post_url
    dtype: string
  - name: topic_model_medium
    dtype: string
  - name: topic_model_broad
    dtype: string
  splits:
  - name: train
    num_bytes: 1959958888
    num_examples: 97033
  download_size: 1928991796
  dataset_size: 1959958888
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# OpenAI Community Posts

This dataset is curated from the posts of the OpenAI Community Forum (https://community.openai.com).

![image/png](https://cdn-uploads.huggingface.co/production/uploads/64a3efa56866210ffc6f83f1/C7FF2hLRHO6A-PDxni-Dw.png)

## Dataset Details

### Dataset Description

The OpenAI Community Posts dataset comprises discussions, posts, and metadata from the OpenAI Community Forum. 
It includes details such as discussion titles, tags, views, reply counts, post content, sentiment scores, vector embeddings for content analysis, and identifiers linking posts to discussions. 
The dataset aims to facilitate analysis on community engagement, content sentiment, and discussion dynamics.

_The dataset includes post from the creation of the forum till Feb 28th, 2024_

The dataset was primarily gathered to understand the sentiment of different OpenAI products amongst the users as well as to gather feedback, complaints and common problems users faced.

Posts from the following [categories](https://community.openai.com/categories) and their relevant sub-categories are included:
- [API](https://community.openai.com/c/api/7)
  - API/Bugs
  - API/Deprecations
  - API/Feedback
- [GPT Builders](https://community.openai.com/c/gpts-builders/33)
  - GPT Builders/Chat-Plugins
  - GPT Builders/Plugin-Store
- [Prompting](https://community.openai.com/c/prompting/8)
- [Community](https://community.openai.com/c/community/21)
- [Documentation](https://community.openai.com/c/documentation/14)

- **Curated by:** Julep AI 
- **Language(s) (NLP):** English

### Dataset Sources [optional]

- **Forum:** https://community.openai.com

---

## Dataset Structure

The OpenAI Community Posts dataset is structured around two primary entities: discussions and posts. Each discussion comprises multiple posts, including an initiating post and subsequent replies. 
The dataset includes various features capturing the characteristics and metrics of both discussions and posts, as well as sentiment analyses and vector embeddings for advanced content analysis.

### Fields Description

- **Discussion-Level Features**:
  - `post_discussion_id`: Unique identifier for the discussion.
  - `post_discussion_tags`: Tags or keywords associated with the discussion.
  - `post_discussion_title`: Title of the discussion.
  - `post_discussion_created_at`: Timestamp indicating when the discussion was created.
  - `post_category_id`: Identifier for the category under which the discussion falls.
  - `post_discussion_views`: Number of views the discussion has received.
  - `post_discussion_reply_count`: Count of replies or posts within the discussion.
  - `post_discussion_like_count`: Number of likes the discussion has accumulated.
  - `post_discussion_participant_count`: Number of unique participants in the discussion.
  - `post_discussion_word_count`: Total word count of all posts within the discussion.
  - `post_discussion_related_topics`: Related topics or discussions.
  - `post_discussion_url`: Web URL of the discussion.

- **Post-Level Features**:
  - `post_id`: Unique identifier for the post.
  - `post_author`: Name or identifier of the post's author.
  - `post_created_at`: Timestamp indicating when the post was created.
  - `post_content`: HTML content of the post.
  - `post_read_count`: Number of times the post has been read.
  - `post_reply_count`: Number of replies to the post.
  - `post_author_id`: Unique identifier for the post's author.
  - `post_number`: Sequential number of the post within the discussion.
  - `accepted_answer_post`: Boolean indicating if the post is marked as the accepted answer to the discussion.
  - `post_content_raw`: Markdown formatted content of the post.
  - `post_category_name`: Name of the category to which the post/discussion belongs.
  - `post_sentiment`: Sentiment of the post (e.g., positive, negative, neutral).
  - `post_sentiment_score`: Numerical score representing the sentiment of the post.
  - `post_content_cluster_embedding`: Vector embedding for clustering purposes.
  - `post_content_classification_embedding`: Vector embedding for classification.
  - `post_content_search_document_embedding`: Vector embedding intended for enhancing search functionalities.
  - `post_url`: Web URL of the post.

### Additional Notes

- **Relationships**: Each post is linked to a discussion through `post_discussion_id`, facilitating analyses that require context from the discussion level or aggregations at the discussion level.
- **Vector Embeddings**: The inclusion of vector embeddings (`post_content_cluster_embedding`, `post_content_classification_embedding`, `post_content_search_document_embedding`) enables advanced NLP tasks, including but not limited to clustering, classification, and enhanced search capabilities within the dataset.
- **Sentiment Analysis**: Sentiment scores (`post_sentiment`, `post_sentiment_score`) provide insights into the emotional tone of posts, useful for content analysis, community mood tracking, and identifying discussions that may require moderator attention.

This structure supports a wide range of analyses, from basic statistical summaries to complex machine learning models, by providing comprehensive metadata, content, and derived metrics for each post and discussion in the OpenAI Community Forum.

## Dataset Creation


### Curation Rationale
The OpenAI Community Posts dataset consists of discussions and posts from the OpenAI Community Forum, specifically curated to analyze developer sentiment, identify common problems, and gather feedback on OpenAI products. It includes detailed metadata for discussions and posts, sentiment scores, and vector embeddings for content, facilitating a comprehensive analysis of community engagement and response to OpenAI's offerings. This dataset serves as a valuable resource for understanding the needs, challenges, and perceptions of developers using OpenAI technologies, contributing to product improvement and community support.


#### Personal and Sensitive Information
Efforts were made to anonymize personal information where possible, excluding direct identifiers but including publicly shared content and metadata for analysis.
Specifically, `post_author` field was dropped and `post_author_id` was converted to a SHA256 hash to preserve user identification.