Start your environment by clicking Start Lab above.

The environment should begin to load immediately as indicated by the Vocareum loading symbol. Please do not click Start Lab again. It may take a few minutes for the Fusion environment to fully display.

When the Fusion Login page displays, login:
USERNAME: admin
PASSWORD: password123

Welcome to the Natural Language Processing Lab. In this lab you will learn how to use deploy a spaCy model, create a machine learning job, and add fields to the Query Workbench. Let's get started!

Note: You may need to adjust the size of the instructions panel to fully view results in the UI. Toggle the window sizes as needed.

  1. Click on the app Labs to enter the Fusion workspace

  2. Hover over the COLLECTIONS icon and click Jobs

  3. Click Add+

  4. Begin typing “create seldon” and select Create Seldon Core Model Deployment from the dropdown

  5. In the Create Seldon Core Model Deployment window enter the following info for the fields shown below:

    a. JOB ID: Labs-spacy

    b. MODEL NAME: spacy

    c. DOCKER REPOSITORY: lucidworks

    d. IMAGE NAME: spacy-seldon:v1.0

    e. OUTPUT COLUMN NAMES FOR MODEL:
    [token_offsets, pos_labels, ner_offsets, ner_labels, lemma_offsets, sentence_offsets]


  1. Click Save

  2. Click Run, then click Start

Note: Success! The job will take about a minute, when it's done, the "running" icon will change to a "sunshine".

  1. Click Save in the Run Job window

  2. Hover over the INDEXING icon and click Index Pipelines

  3. Click the Add+ button

  4. For Pipeline ID, enter labs-spacy

  5. Click Add a new pipeline stage

  6. Begin typing "machine" and select Machine Learning from the dropdown

  7. In the Machine Learning window enter the following info for the fields shown below:

    a. Label: Extract Entities

    b. Model ID: select spacy from the dropdown

  8. Click into the field for Model input transformation script

  9. Scroll to line 46 of the code. Cut and paste the */ before line 42, as shown below:


Note: This will "comment out" everything before line 15, making it inactive

  1. Replace line 44 of the script with the following:

modelInput.put("input", doc.getFieldValues("longDescription"))

Your script should resemble the image below:


  1. Once you are finished modifying the script, close the script window

  2. Returning to the Machine Learning window, click into the field for Model output transformation script

  3. Scroll to line 16 of the code. Cut and paste the */ before line 14, as shown below:


  1. Remove the existing text after the */, and replace it with the following code:

doc.addField("ner_labels_ss", modelOutput.get("ner_labels")) doc.addField("ner_offsets_ss", modelOutput.get("ner_offsets")) doc.addField("pos_offsets_ss", modelOutput.get("pos_offsets")) doc.addField("pos_labels_ss", modelOutput.get("pos_labels")) doc.addField("lemma_offsets_ss", modelOutput.get("lemma_offsets")) doc.addField("sentence_offsets_ss", modelOutput.get("sentence_offsets"))


Note: This is how we will create new fields to appear in the Query Workbench!


  1. Your Model output transformaton script should now resemble the following:

  1. Once you are finished modifying the script, close the script window

  2. Click the Save button in the upper right corner of the Machine Learning Pipeline window

  3. Hover over the COLLECTIONS icon and click Jobs

  4. Click the job BestBuy_catalog

  5. Under the READ OPTIONS section, in the SEND TO INDEX PIPELINE field, enter labs-spacy


  1. Click the Save button. Then, click Run, and Start.

    a. Note that this could take up to 5 minutes to completely run. Once the job status indicates success, click Save

  2. Hover over QUERYING and click Query Workbench

  3. Under the first result, click show fields

  4. When we scroll down a little, we can now see the fields we added in the Index Pipeline


Now, let's add some field facets using our new fields

  1. In the Query Workbench window, click Add a field facet

  2. Begin typing "pos" and select pos_labels_ss

The field facet then shows the different parts-of-speech labels that are in the long descriptions of each product


  1. Click Add a field facet again, and begin typing "lemma". Select lemma_offsets_ss

This field facet organizes when, and how many times, a person, quantity, organization, time (and much more) are mentioned in the long descriptions of each product

Make sure to Save your open Fusion Workspace tabs!

Great job! You have successfully explored how to use NLP capabilities in Fusion! If you would like to save your Fusion App to reference later, you can do it now:

  1. Return to the Fusion Launcher
  2. Hover over your app and click on the cog that appears in the lower right corner
  3. Within the box that opens, click Export app to zip

Click Next in the lower right corner of the screen to continue on in the course