The Kore.ai platform provides a Taxonomy-based model in the Knowledge Graph to enhance the path qualification.

The default Knowledge Graph model works on a two-step model i.e. path qualification and followed by question matching. The path need not be fully qualified at all times. Even a partial path match (above a threshold) is considered as a qualification and the questions in these paths are used for matching the user input.

In the ‘taxonomy’ based approach, the ‘path’ should fully match at all times. The idea here is that every term in the path is equally important and only a full match of all the terms in the path should be considered as a qualification. Once a path is qualified, the questions in that path or paths will be considered for intent identification against user input. This is useful when you want to ensure that FAQs are answered from a given path only when the complete path is either present in the user’s input or disambiguated from the user to answer the most confident questions.

To achieve this, the platform allows a custom configuration settings Taxonomy-based Knowledge Graph under the Advanced NLP Configurations. By default, this setting is disabled. When enabled, the Knowledge Graph evaluates FAQs using the taxonomy-based approach.

This feature was released with v9.0 of the platform and is in (beta) state.

Use Case

Consider the following Knowledge Graph:

Any query related to accounts – opening or closing would yield in the following:

With Taxonomy enabled KG, even partial match of the node/terms will yield in a response that is useful:

Advantages

1. Taxonomy-based KG helps disambiguate at the parent level. Nodes/terms are considered more important in the KG qualification.
2. It allows for multiple questions within one leaf by allowing you to link existing questions to new intents.
3. Allows you to give display names for nodes for presenting to the user.

Enabling

To enable the Taxonomy-based custom configuration settings, follow the below steps:

1. Log in to the Kore.ai platform with valid credentials.
2. By default, the Bot Summary page is displayed.
3. On the left pane, click Natural Language -> Thresholds & Configurations.
4. On the Thresholds & Configurations page, click Advanced NLP Configurations. You can configure the thresholds and configurations associated with the advanced NLP settings.
5. In the Select a Configuration drop-down field, select Add Custom from the list.
6. In the Custom Configuration field, type KG_Taxonomy_Based and click Enter.
7. In the Configuration Value field, type enable and click Enter to save the configuration.
   

Features

Once you enable the Taxonomy-based KG, the following options are enabled:

1. For each node/term, you can:
   • enter the Term Display Name – this is presented to the user for disambiguation
   • enable the Auto Qualify Path – this will ensure that the term is auto-qualified if any of its immediate child terms are qualified.
     
2. When adding a new intent, you have the option to Link an Existing Question – this might  prove useful since the user might provide a partial input that contains one or more terms
   

Notes

Once you enable the Taxonomy-based KG, keep in mind the following:

1. For the Ranking and Resolver engine the following settings would be disabled:
   • Prefer Definitive Matches, and
   • Rescoring of Intents
2. The following settings of the Knowledge Graph engine would be ignored:
   • Path Coverage,
   • KG Suggestions Count, and
   • Proximity of Suggested Matches.