This feature is currently in an open beta and opt-out only. Visit the section below for instructions on how to opt-out of AI features.
In this article:
About Iris post analysis (powered by OpenAI ChatGPT)
Iris is Brandwatch’s AI. It uses cutting edge proprietary and integrated AI models to help you work faster and smarter. Iris post analysis leverages OpenAI GPT (Generative Pre-trained Transformer) technology to provide an analysis of post content and comments in Measure’s content and label performance widgets to give you quicker insights about your content.
Generative Pre-trained Transformer technology is an advanced language model developed by OpenAI. It utilizes deep learning techniques and a vast amount of training data to understand content and generate insights. Visit Open AI’s website to learn more about OpenAI and ChatGPT.
Using Iris post analysis
You can find Iris post analysis in Measure when viewing the Post Details of a post from a Content or Label Performance widget. Click the Analyze Post button inside the Post Details dialog box to analyze a post and its comments.
After clicking the button, you will see the Iris post analysis tab with an indication that Iris is analyzing the post’s sentiment.
If the post has no post text to analyze, the post analysis tab will display the message, “There’s no text to analyze in this post.” If the post has no comments, the tab will display the message, “There’s no comments on this post yet.”
If there was a general error processing the information of the post, the tab will return the error message, “Sorry, we’re not able to analyze this post right now. Please try again later.”
When the data is available, the tab will show two sections: Post analysis and Comment analysis.
Post analysis analyzes the message/text of the post. The data will be returned as a table with the following columns:
- Style: The structure of the wording of the post.
- Complexity: How easy your post is to read or understand.
- Sentiment: The feeling or emotion of the post.
- Tone of voice: The tone in which the post is communicated.
- Links: Any links included in the text.
- Hashtags: Any hashtags included in the text.
Comment analysis analyzes the comments of the post and displays the top five topics discussed by the 50 most recent comments. The data will be returned as a table with the following columns:
Replies to comments are not included in comment analysis.
- Top Themes: The main topics of conversation in the comments.
- Number of comments: How many comments your audience has published about the specific topic.
- Sentiment: The feeling or emotion of the topic.
- Tone of voice: The overall tone of the comments.
- Style: The structure or wording of the comments.
Opting out of AI features
- In the left-hand navigation sidebar of Social Media Management, click on Manage Settings (gear icon), then click Organization admin.
- Under OpenAI features, switch the Enabled toggle on the right side of the screen to Disabled.
This setting will disable all AI features Social Media Management. You can return to this menu to re-enable your AI features at any time. If you have access to both Social Media Management and Consumer Research and would like to opt out of Consumer Research’s Iris Conversation Insights, you will need to opt out separately via your Consumer Research settings.
Which current Brandwatch features use GPT technology?
AI-search and AI-topics in Consumer Research are both powered by AI trained using GPT-2. GPT-2 is an older large language model (LLM) by OpenAI.
How does data sharing with OpenAI vary by Brandwatch’s AI features?
Iris Conversation Insights, Iris Writing Assistant, and Iris Content Insights each send data directly to OpenAI. This will not include any user data, only data within the tools themselves. So in the case of Iris Writing Assistant, this could be any text you draft. We always require that OpenAI does not use this data for training purposes.
Regarding Iris Conversation Insights, Consumer Research shares snippets of text from a sample of mentions with OpenAI. This sample is currently generated randomly. Image data found in sample mentions is currently not shared. Additionally, we do not send any query data (e.g. the keywords). There are no immediate plans to share anything other than mention content with OpenAI.
Regarding sensitive information found in the mentions themselves:
- We ask clients to be aware that all data found within the mentions panel of Consumer Research could be shared with OpenAI (if you use an OpenAI feature). This includes uploaded data sets.
- If those mentions contain PII as defined by the client, these could also be shared to OpenAI to derive summaries.
- However, OpenAI will not use it for their own purposes and will only store it for 30 days for support/security purposes.
AI-search and AI-topics were developed in-house and are not direct integrations with OpenAI. We do not send any data outside of Brandwatch with these features.They use an earlier version of GPT (GPT-2), not ChatGPT.
Will these features work for different languages?
Iris Writing Assistant works in the languages that the user uses while writing. It supports all languages that ChatGPT supports.