Get Started with Fusion Server:
Improve Relevancy

Signals are events that can be aggregated and used for automatic boosting or recommendations, which are ways of making search results more relevant.

As an example of boosting, the most popular search results for certain queries can be boosted so that they appear first (or at least nearer the top) when other users make similar queries.

Similarly for recommendations, click events or purchase events can be collected as signals and used to display "Customers who viewed this also viewed" or "Best-selling holiday items".

In a production environment, users' actions generate signals. For the purposes of this tutorial, you’ll use Query Workbench to generate click signals.

Before you begin

To proceed with this part of the tutorial, you must first complete Part 1, Part 2, and Part 3, which give you an indexed dataset that’s configured for faceted search.

If the Fusion UI isn’t already open, then open it.

  1. In a browser window, open localhost:8764.

  2. Enter the password for the user admin, and then click Login.

    The Fusion launcher appears.

  3. Click the Movie Search app.

    The Fusion workspace appears.

1. Format display fields

To help you understand the continuity in the next steps, first make sure that some relevant fields are displayed.

  1. Open Query Workbench. Navigate to Querying Querying > Query Workbench.

  2. At the top right of the page, select Display Fields.

  3. In the Name field, click the row to reveal a dropdown of possible values.

  4. Select title_txt. You can filter the list of possible values to help you.

  5. In the Description field, select id.

  6. Click Display Fields to close the Display Fields window.

    Query Workbench before starting Part 4 of the Getting Started tutorial

2. Enable synthetic signals

You need some signal data beyond the few signals you generated while completing Part 3. Because this is a prototype app, users aren’t generating signals. Instead, you’ll enable synthetic signals in Query Workbench.

  1. At the bottom of the Query Workbench page, click Format Results.

  2. Select Show signal generators and Send click signals.

    Configure signals

  3. Click Save.

  4. Hover over one of the search results.

    Now when you hover over a search result, Query Workbench displays controls that include a Simulate button next to a field that lets you specify the number of signals to simulate:

    Controls to simulate signals

3. Generate signals

With synthetic signals enabled, you’ll generate a simple set of signal data that you can use to generate meaningful recommendations.

For this tutorial, you’ll generate signals that you can use to boost your favorite sci-fi titles so that they appear first.

  1. Search for star wars.

    The top results are not your favorite titles:

    Star Wars search results

    Next you’ll generate signals that you can use to boost certain titles. Signals are tied to the search query, so your boosted titles will appear first in the search results only when users search for star wars.

  2. Hover over "Star Wars: Episode IV - A New Hope".

  3. Set the number of signals to 4000 and click Simulate.

  4. Hover over "Star Wars: Episode V - The Empire Strikes Back".

  5. Set the number of signals to 3000 and click Simulate.

  6. Hover over "Star Wars: Episode VI - Return of the Jedi".

  7. Set the number of signals to 2000 and click Simulate.

    Tip
    With synthetic signals enabled, you can also send a single signal by clicking the underlined movie title.
  8. In the upper right, click Save.

    The Save Pipeline window appears. By default, you’ll overwrite the existing query pipeline for the selected collection (in this case, Movie_Search_signals).

  9. Click Save pipeline.

4. Explore the raw signals

Whenever you create a collection, two corresponding collections are also created automatically: <collection>_signals for raw signals and <collection>_signals_aggr for aggregated signals (in this case, Movie_Search_signals and Movie_Search_signals_aggr). Just as you did with your primary collection, you can use Query Workbench to explore the data in the _signals collection.

  1. In the collection picker in the upper left, select Movie_Search_signals.

  2. Open Query Workbench. Navigate to Querying Querying > Query Workbench.

    Your signal data appears.

    Signal data

    Your signals collection contains several types of signals, such as click signals and search result signals. Let’s investigate the click signals.

  3. Search for type:click.

  4. For any of the results, click show fields.

    • The count_i field shows the number of click signals you generated for this event. For example, for 260 (the doc_id_s for Star Wars: Episode IV - A New Hope), count_i equals 4000.

      Number of click signals for document 260

  5. Click hide fields.

4.1. Optional: Format signals fields

You can configure your search view to display fields that are meaningful for your investigation. For example, you can display the document ID and the number of click signals.

  1. Click Display Fields.

  2. For the Name field, select doc_id_s.

  3. For the Description field, select count_i.

    Doc ID and count fields for click signals

5. Explore the aggregated signal data

Aggregation jobs are enabled automatically. Aggregation jobs for click signals run every fifteen minutes by default to aggregate any new raw signals. Aggregations jobs for other types of signals run at other frequencies by default.

Check whether the aggregated data has arrived in the Movie_Search_signals_aggr collection.

  1. In the collections picker in the upper left, select Movie_Search_signals_aggr.

  2. Open Query Workbench. Navigate to Querying Querying > Query Workbench.

    Your aggregated signal data should appear. If not, wait a minute and then reload your browser, or click Search Search in Query Workbench.

    Aggregated signals

  3. Click Display Fields.

  4. For the Name field, select doc_id_s.

  5. For the Description field, select aggr_count_i.

  6. For the result 1210, click show fields.

    Aggregated signal fields

    The fields for aggregated signals are very similar to the fields for raw signals, with additional fields to describe the aggregation:

    • aggr_count_i – Number of signals that have been aggregated (in this case, 3000)

    • aggr_id_s – Name of the aggregation job

    • aggr_job_id_s – Job ID

    • aggr_type_s – Aggregation type (in this case, sql)

  7. In the upper right, click Save.

    The Save Pipeline window appears. By default, you’ll overwrite the existing query pipeline for the selected collection (in this case, Movie_Search).

  8. Click Save pipeline.

6. View the search results with and without default boosting

  1. In the collections picker in the upper left, select Movie_Search.

  2. Open Query Workbench. Navigate to Querying Querying > Query Workbench.

  3. Search for star wars.

    Now, "Star Wars: Episode IV - A New Hope" is the first search result, followed by Episode V and then VI. These search results are automatically boosted by the default configuration of the Boost with Signals query pipeline stage, which boosts on the id field.

    Star wars movies boosted

    Note
    The Boost with Signals stage requires a Fusion AI license. Your Fusion Server trial license enables Fusion AI features.
  4. Click Compare. Another preview panel opens. Now the working pipeline is on the right and a static snapshot of that same pipeline is on the left.

    Compare two pipelines

    In this view, you can compare results from one query pipeline side by side with another query pipeline. In this case, you’ll compare results for the same pipeline (Movie_Search) with and without the Boost with Signals stage enabled.

  5. Turn off the Boost with Signals stage.

    Boosting comparison

    Now the search results on the right appear as they did before you generated synthetic click signals. Rank-change indicators indicate which results moved up or down as a result of turning off boosting.

  6. Turn on the Boost with Signals stage again to restore the boosted results.

  7. Close the comparison preview panel by clicking the close Close icon.