Why some data is ignored by our AI models!
Not all data is processed by all AI services. Certain information is intentionally ignored by our AI models. If you are missing information, feel free to look at the reasons why data is ignored below and see if any of them apply to your situation.
- attachments of any kind
- Text elements that consist of embedded PDFs or IFrames only
- Text elements that contain only email addresses or links
- Text elements that consist of more than 500,000 HTML characters
To have AI services suggest information from attachments or text elements with embedded PDFs, IFrames, email addresses, or link, simply add a brief summary describing what information you can glean from the attachments, embedded PDFs and IFrames, or links.
Typically, your text elements will contain no more than 500,000 HTML characters. For comparison, Thomas Mann's Wikipedia entry contains about 400,000 HTML characters. However, there are some reasons why texts can sometimes contain so many HTML characters without you seeing it directly:
-
Base64 image
If images have not been embedded into the text via Image Explorer (but e.g. via drag & drop), this can lead to a lot of additional HTML characters that are needed to display the image. -
Embedded Word documents
If an entire Word document has been embedded in the text, the HTML code will also be very long. Here's how to recognize an embedded Word document in the HTML code: <w:WordDocument>Welcome to Word Document</WordDocument>.
As a reminder, you can open the HTML code by selecting the "Source" button at the top left of the CKEditor toolbar.
There is another reason why you might miss information in the Contact Deflection Widget (AI Widget) suggestions:
- If the user's view used to create the AI Widget does not match the view of the text, the text cannot be suggested.
If none of the above reasons apply to your situation, feel free to contact our support team!