IBM Cloud Docs
Previewing your query results

Previewing your query results

See the types of query results that are returned automatically and learn about how they are derived.

When a document is ingested, the text is extracted and indexed in thetext field. To return only the subset of information that is relevant to the query, Discovery returns passages from the text field. For more information about passages, see How passages are derived.

When you enter a query from the product user interface, the UI submits the text as a natural language query. For more information about how to query your data programmatically, see Query API.

To query your data from the product user interface, complete the following steps:

  1. From the navigation panel, open the Improve and customize page.

  2. Take the appropriate next steps for your project type.

Document Retrieval

  1. Do one of the following things:

    • Click Run search for one of the keywords that Discovery calculated to have special meaning in your collection.
    • Submit your own phrase or keyword from the search bar.

    You can see that the query results that are returned consist of passages.

    Entities that are recognized in your documents (based on the Entities enrichment that is applied to the project by default) are displayed as facets by which you can filter the query results.

  2. To explore a query result in more detail, click View passage in document.

  3. Click Open advanced view to explore the entity mentions that are recognized by Discovery.

Excerpt unavailable

Passages displayed in search results are extracted from the content that is indexed in the title and text fields of the documents. If the content is indexed in other fields, the search displays the Excerpt unavailable message.

Content is indexed in other fields in the following scenarios:

  • Your collection contains structured files, such as JSON or CSV files. When you ingest structured files, content is stored in custom fields with names taken from the original object names (JSON) or column headers (CSV).

  • You applied a Smart Document Understanding model that moves content from the text field into new fields, such as section or results, based on the document's structure.

To improve the search results, first choose how you want to extract content from the documents. If your documents have targeted fields with succinct content in them, such as answer fields for an FAQ use case, configure the search to return those specific fields. If your documents have custom fields with lots of content in them, such as chapter fields, configure the search to find the best passages from the custom fields.

To customize the results, complete the following steps:

  1. From the Improvement tools panel, expand Customize display.

  2. Click Search results.

  3. In the Select source of result content option, do one of the following things:

    • Select Passages, and then specify one or more fields from which to extract passages.

    • Select Field, and then choose one or more fields to return.

    Shows the Search results dialog
    Search results dialog

  4. Click Apply.

Conversational Search

A single search field is displayed that mimics the user interface of a virtual assistant.

  1. Submit a phrase or keyword.

The query results are returned as passages by default. You can configure the search to return a specific field instead.

If you want to investigate the results a bit more, you might want to use a different project type. For more information, see Improving results for a chatbot.

Document Retrieval for Contracts

Contract-related elements that are recognized in your collection are displayed.

  1. Do one of the following things:

    • Filter the documents by one of the highlighted elements or by entities that are recognized in your documents (based on the Entities enrichment that is applied to the project by default).
    • To view the contract elements in more detail, click a document result to open it. Open the Contract Data tab.

For more information about the elements, see Understanding contracts.

Content Mining

Facets based on the Part of Speech enrichment are shown.

  1. To analyze your data, open the Content Mining application. Click Launch application.

For more information, see Analyzing your data.

What to do next