Getting started with the Oracle Analytics AI Assistant
Science & Technology
Introduction
In this article, we'll walk through the fundamental steps required to begin using the Oracle Analytics AI Assistant, a powerful tool for analyzing data efficiently.
Loading Data into the System
To get started, the first task is to load some data into the system. For this demonstration, I'll simply drag and drop a spreadsheet onto my homepage. This spreadsheet contains general information about sales, customers, and products—essentially basic datasets.
Once loaded, I'll navigate to the "Data" tab. The next step is to access the "Inspector" and then the "Search" tab. Here, I can enable the dataset to be available for the assistant. When this option is selected, a list of columns, attributes, and measures available for the assistant to search through will be displayed.
Configuring Columns for the Assistant
For each column that should not be indexed by the assistant—like order line IDs or technical product keys—I can easily exclude them from the configuration. This ensures the assistant focuses only on the relevant columns, such as order IDs, order priority, customer segments, and product categories.
As I scroll through the list of columns, it’s crucial to be as descriptive as possible. For instance, "Customer ID" could be referred to as "Customer Number" or "Customer Identifier." This descriptive approach assists the language model in understanding the data better. Thus, adding synonyms is encouraged, as you can include up to 50 synonyms per column.
Another helpful feature of the synonyms interface is that it retains the definitions across different datasets. Consequently, if a new dataset shares similar columns, it will suggest previously defined synonyms, saving time on future configurations.
Once the synonyms are set, I can opt to save the dataset or send the information to the language model by clicking "Run Now."
Interacting with the Assistant
Now that the data is configured, I can create a workbook from my dataset. By clicking the light bulb icon in the top right corner, I will find the assistant tab, and I'll check the status of information being sent to the language model. This process can take a few minutes, but typically, it completes swiftly.
Once ready, I can begin querying the data. A simple initial query might be, "What were my sales by segment and state?" The results will present my customer segments aggregated by each state, which I can then drag onto my canvas for deeper analysis.
I can also try a follow-up question like, "What was my profitability by category for the last three months?" The assistant filters the data based on the specified time frame, effectively narrowing down the results to a manageable dataset.
However, not all queries will succeed. For example, if I ask, "How many corporate customers placed orders this year?" the assistant may struggle and return no results. This often means the assistant lacks an understanding of certain terms—such as "corporate customers" in this case. A more straightforward question, like "How many customers placed orders this year?" is a better approach to retrieve relevant information.
In summary, working with the Oracle Analytics AI Assistant involves loading data, configuring data properties for assistant queries, and efficiently interacting with the assistant to extract meaningful insights.
Keywords
- Oracle Analytics
- AI Assistant
- Data Loading
- Inspector
- Synonyms
- Configuring Data
- Customer Segments
- Queries
FAQ
Q: How do I load data into the Oracle Analytics AI Assistant?
A: You can load data by dragging and dropping a spreadsheet onto your homepage within the Oracle Analytics interface.
Q: What should I exclude while configuring the dataset for the assistant?
A: Exclude technical data such as order line IDs or product keys that are not relevant for querying.
Q: How many synonyms can I add per column?
A: You can include up to 50 synonyms for each column in your dataset to improve the language model's understanding.
Q: How does the assistant learn from the synonyms I set?
A: The assistant remembers the synonyms you define across datasets, making it easier for future datasets to use similar terms without re-entering them.
Q: What to do if the assistant doesn’t understand my question?
A: If the assistant fails to understand, consider rephrasing your question using simpler or more universally recognized terms.