- Run iterative natural language queries on a DocSet
- Retrieve the results of a query, which can include text, tables, and lists of relevant documents
- Save the results of a query as a Bookmark, and query the Bookmark
- Adjust the query plan in natural language before running it
- View the input and output into each query step (for debugging purposes)
Running a query
Go to the Workspaces UI by clicking on it in the left nav, and then select the DocSet you want to query. Or, go to your DocSet in the DocSet list page and click the Open Workspace button. Once you’ve selected your DocSet in the Workspace UI, enter a natural language query in the query box and choose the execution mode using theDeep Analytics
toggle. The bottom of the box has the Dataset selector, which defaults to the DocSet you selected for your Workspace. If you create a Bookmark in your Workspace (see below), you can select it here to query it.
After you submit your query, Aryn will generate and show you the query plan, or steps it will take to execute your query based on the Deep Analytics
setting. You can choose to run the plan by clicking the Run plan button, or cancel the query by clicking the Cancel query button. This will allow you to submit a new query.
Editing query plans
If you need to adjust the query plan, select the Edit plan button. You can instruct Aryn to change the plan in natural language using the query box. Submit your instructions, and you should then see the query plan change. Once you are satisfied with the plan, click the Run plan button. Note that you can not edit the plan whenDeep Analytics
setting is not enabled.
Query results
Aryn returns a result and a trace from your query.Query result
Depending on the query, Aryn will return results in different formats:- List of documents: The output of a query can be a new DocSet. For example, a query could filter a DocSet and return the filtered set.
- Table: The output of a query can be a table with derived values, like counts.
- String: The output of a query can be a natural lanauge answer, like a summary or question-answering text.