Report Building with AI
Last updated
Last updated
On every page of the universe, you can leverage the help of AI to write your report, entity page, etc.
You can access the AI-report generation menu called Ask IGOR^AI in various ways. Most noticeably, when editing a page, you will find the Ask IGOR^AI button on the top left of the toolbar.
Alternatively, Ask IGOR^AI can also be accessed in the editor when inserting and selecting content:
When adding a new line, hit /
(forward slash) and a menu appears. Ask IGOR^AI will be the first item in the list.
When hovering over the content blocks, you will see a +
(plus) sign appear on the left side. When clicking it, a menu appears where Ask IGOR^AI is the first item in the list.
When selecting text, the same toolbar appears as on the top of the editor, with again Ask IGOR^AI on the left side.
Hereunder a video outlining the different ways to access Ask IGOR^AI:
This will lead to a pop-up in which you can ask IGOR^AI to do various tasks based on your documents such as:
Question & Answer: This can be used to generate specific report parts for you or to ask questions about the documents you saved
Report: Writes a report with an Introduction, Methods, Results, and Conclusion
General description: Writes a general description of your entity or study
Write section: Writes a lengthy section with a header of your choice
Extract detail information: Extracts requirements from every paper saved to the entity or study
Executive summary: Writes an executive summary based on all the information you have added
Ask IGOR^AI is constrained to look only in the documents and highlights saved to an entity or study to provide its answers. This means it will not provide an answer if that answer is not findable in the documents and highlights saved.
The AI will search in two parts of the documents you saved:
Highlights you saved;
For papers, it will also look into the paper abstract;
For patents, it will look at the abstract and the claims.
The highlights you saved on each document are prioritized in Ask IGOR^AI's response.
Every task has a specific set of sources from which to extract information and generate content.
Based on the selected amount, it will scan all abstracts/descriptions/claims, highlights of the documents, and highlights linked to the entity or study for relevancy on the subject(s) you prompted, rank them, and create a selection to generate content with.
Hereunder, the standart
Question & Answer
-
Top 10 + top 10 highlights
Top 10
Report
-
Top 10 + top 10 highlights
-
General description
Yes
Top 10 + top 10 highlights
Top 10
Write section
-
Top 10 + top 10 highlights
Top 10
Extract detail information
-
All documents (no highlights)
-
Executive summary
Yes
Top 10 + top 10 highlights
Top 10
This is the text that is on the page itself where you want IGOR to generate new content for.
The information of the documents used to generate content differs per type of document:
Science: abstract
Patents: claims and abstract
Web: site description
You can directly link highlights to a page even if the document, where it came from, isn't connected.
Note: Findest isn't able to read any content outside of the Universe. Saving a webpage doesn't mean that the content is automatically saved with it. Use highlights to add important information from the page that IGOR^AI can use to generate content.
There are multiple ways to have more control over the sources used by Ask IGOR^AI to generate content with or extract data from.
By default, the amount of sources is set to 10, as it has the perfect balance between speed and accuracy. However, there are numerous cases in which you would want to either increase or decrease this number. By clicking on the sources displayed underneath the description of the task, a settings pop-over will appear.
Use the slider or the number selector to set the maximum number of sources that you want to use. The range goes from 3 to All. If you have 7 documents linked to a page and want them all to be used, you can leave the amount at 10, as you are setting a limit for the maximum amount of sources used.
In this same settings menu, you can also select what type(s) of documents you want to filter for Ask IGOR^AI to use. By default all types are selected, but if you only want to include science documents in your generated content, simply deselect the other types (Web and Patents) so that Science is the only type selected.
Hereunder a video exemplifying the steps taken to adjust the sources and the type of documents:
If you wish to only include specific documents as source for your generated content, you can now do that by going to the "Linked documents" section near the bottom of your page. Here you can create a selection of documents by clicking on the checkboxes, and click the Ask IGOR^AI button that appears next to the link/unlink buttons on the top of the documents list.
When choosing a task, you will see that underneath every description it is indicated that a number of selected documents is going to be used, and the usual settings are not available.
There is a limit to the amount of tokens we can send to the language model. In general, the limit that we put on the amount of documents and highlights used will strike a great balance between speed and accuracy. However, we can't control the size of the highlights that you create.
As there is no limit on how big a highlight can be, this might become a problem when having to process such large amounts of text. You might run into the following issues:
It takes a very long time to process, rank and generate information.
The generated content is not what you expected as the large amounts of texts the model had to go through were not focused.
You'll get an error if the amount of text exceeds a certain limit.
Guide IGOR^AI by only highlighting information that is relevant to your project. More data doesn't always mean the retrieval quality is also able to catch the information that you want.
The and allow you to control the output depending on your needs. They are investigated more thoroughly in the respective sections.
By default the number of sources is set to 10 to have the perfect balance between speed and accuracy, but you can customize this by going to .
+
(plus) next to a content block in the editor.