Skip to main content

Configuring the Recommender - Preliminary Steps

Abstract

Configuring the Recommender - Preliminary Steps

This section will give you some initial insights regarding the recommender configuration including selection of

  • the Elasticsearch index

  • the desired corpus and search space

The key to a knowledge based recommendation system is transparent operation and reproducible and verifiable results. The subject matter expert configures the system with the Recommender Workbench as the tool used to set the parameters for the extraction process, then the rules for the semantic expansion and finally the recommendation. The Recommender Workbench provides the expert with a tool assisting them in thorough verification of the settings.

Let us take a closer look at the requirements to be satisfied before we can use the Recommender Workbench.

Tip

You need to set the correct indexing in the Semantic Middleware Configurator.

Elasticsearch Index

The PoolParty Suite supports Solr and Elasticsearch indices. However the recommendation calls explicitly require Elasticsearch index to be in place. To make sure that this is the case, log in to PoolParty as Superadmin and click the Semantic Middleware Configurator icon. On the left navigation tree you see the item Indices - click on it to display the configured index. If an Elasticsearch index is already in place, no interactions are required. Otherwise you need to set it up.

semanMiddleware_elastic.png

Refer to How to Setup a Connection to an Elasticsearch Index if you need to set up an Elasticsearch index.

The PoolParty Semantic Recommender synergistically combines the advantages of concept-based search with the full-text search functionality of Elasticsearch. Elasticsearch index is required for the last call, the recommendation.

Note

Keep in mind that you have to restart PoolParty after changing the index so that the configuration changes can come into effect.

The Search Space

Before configuring the Recommender you also need to specify the search space. This step is done in GraphSearch. To be able to set up a search space you need administrator rights accessing GraphSearch. On the right hand side click the right most symbol to open the menu where you can launch the Admin Dashboard.

searchspace-admin.png

There you see the already set up search spaces with the currently selected one highlighted by a blue background. You can use one of the existing search spaces or create a new one with a click on the "+" symbol.

searchspace_9_1.png

Clicking on the "+" symbol opens a new window where you need to specify the name, type, server, and facet model; optionally you can specify the search space integration you want to use.

You can also view and if necessary edit settings of an existing search space. To do so click Configuration to open the details of the selected search space. You also need to view this configuration in order to copy the Search Space ID value to clipboard. We will need it setting up the recommender.

Note

Make sure that the Selected Server Type is your Elasticsearch server.

Tip

Remember to copy the Search Space ID value to clipboard since you will need this value when writing your recommender configuration.

After setting up the Elasticsearch index and configuring the search space you want to use, you need to make sure that the search space contains the corpus, i.e. indexed documents which are needed for the Recommender to function as intended.

Now you are ready to launch the Recommender Workbench (for details refer to Access the Recommender Workbench). The next section will give you a brief overview of the PoolParty Recommender Workbench functionality while configuring your recommender.