NotebookLM and Clean Core
In this article, we'll explore NotebookLM's features and related documentation, focusing on Clean Core. How can the tool support your learning and provide information?
Table of contents
In this article, I would like to introduce you to a topic that might help you answer your most important questions in the future when researching and gathering information.
Introduction
After completing the article for the new Clean Core Level Concept, I came across the topic of NotebookLM while reading the current news. After reading a few articles and watching YouTube videos, I wanted to use the application for the topic of Clean Core and how it can help you quickly access the most important information.
NotebookLM
NotebookLM is an AI product from Google that lets you create notebooks. In these books, you collect information on a topic. You can integrate various resources from different sources:
- Files - Various types of files, such as PDF documents, audio files, or Markdown files, for example, if you're on GitHub.
- Google - Various documents from Google Docs or presentations.
- Links - If you've found various good articles on a topic online, you can provide the link directly.
- Text - If you've copied text from the document or received it from various emails, you can add these as well.
The various sources serve as the basis of knowledge for the language model. This allows you to work with the model, and it uses the information to answer your questions or create a summary. Unlike a general LLM, the rate of hallucinations is much lower because you work with documents. At the end of the answers, you will also receive notes indicating in which document and in which section the information was found.
Example
As an example, we will use SAP's Clean Core concept. Here, we have three documents on the topics of extensibility, ABAP Cloud, and Clean Core.
Create
To do this, we first create a new notebook in the application. If you already have a Google account, you don't even need to log in to start and use the application. After you have confirmed everything, you will be in the application lobby.
Clicking "Create New" or "Create New Notebook" creates a new notebook. Generally, I would recommend creating one book per topic and adding documents related to the topic and contributing additional content.
To do this, we download the three documents and specify them as the source. After uploading, the notebook will be created and the documents processed. Once the loading process (behind the file) is complete, you can interact with the documents.
Structure
In the upper left area, you can name the notebook. On the left side, you will find your sources, which you can deactivate or activate, as well as add additional sources and information. In the chat, you will see an initial summary of the overall content of your sources. On the right side, you'll find the Studio, where you can generate various content using LLM, such as a podcast, a video, a mind map, or various evaluations.
Studio
In the first step, we'll use the various functions from the Studio to obtain content and a summary.
AI Podcast
In the first step, let's create an audio overview in the form of an AI podcast. A summary is created, and in the style of a podcast, two hosts discuss the topic from the document. We can adjust various settings using the three dots behind the button. We want a short version of the summary, and the focus should be primarily on the Clean Core Level Concept so that we can obtain more information about the topic.
The generation will then take a few minutes before you find the result in the lower area of the studio. Once the generation is complete, you can listen to the podcast. Interactive mode is also available after the generation. This gives you the "raise hand" function while listening; the hosts will respond to you and you can ask questions. The two will then try to answer the question and give you an answer based on the sources.
We've downloaded the resulting audio, and you can listen to it here. The quality is generally very high and is very reminiscent of a podcast.
Video
Let's take a look at the video function and this time generate the video without a specific focus. Simply click the button and the video generation begins. Here, too, the generation process can take a few minutes.
The speaker's animations are quite well designed, and it's a nice summary of the topic. You can see the result of the generation here:
Mind Map
The mind map attempts to represent the various topics and areas in a single mind map. This can be very large, but you can then navigate to the different areas of Clean Core that interest you.
Reports
Under Reports, you have access to various documents that you can generate, such as a short briefing on the topic, an FAQ, learning materials, or a timeline. Basically, the tool can help you get into the topic.
Chat
Finally, let's look at the actual function of the tool: interacting with sources via chat. In this area, you can ask specific questions if something interests you or isn't yet entirely clear. This saves you from having to search through the documents to find the actual answer. So let's start with a question:
- How can we reach Level B in the Level Concept and what do we have to do for that?
After a short period of reflection, we receive the answer to our question; here is an excerpt from the rather long answer.
If you hover your mouse over the number, an excerpt of the reference will be displayed. If you click on the number, the document will open on the left and scroll to the appropriate position. This allows us to always understand the information displayed and continue reading at the appropriate point. Finally, let's summarize the answer in 5 sentences.
Conclusion
When it comes to processing texts and information, LLMs are truly invaluable for saving time and energy. Quick summaries, searches, and source references help you draw the right conclusions. With various formats, such as podcasts, videos, and guides, there should be something for every learning style.
