Create KG from CSV and JSON

Back to NLP Topics

Build your Knowledge Graph in a CSV spreadsheet or JSON file, then import it into your AI Agent. This is useful when your FAQ data already exists outside the platform or when you need to make frequent bulk updates.

Steps:

1. Download the sample CSV or JSON file from the AI Agent.
2. Edit the file to add questions, responses, synonyms, etc.
3. Import the file. See Import a Knowledge Graph.

---

CSV File

Download the Sample

1. Go to Automation AI > Knowledge AI > FAQs > ⋯ (more) > Import.
2. Back up your existing KG when prompted (choose CSV or JSON format).
3. Click Proceed, then click Sample CSV to download.

CSV Structure

The CSV contains five section types, each preceded by a header row.

FAQ Section

Column	Description
Faq	Section header; leave blank in data rows.
Que ID	Auto-generated by the platform. Leave blank for new FAQs; don’t edit for existing ones.
Path	Node path for the FAQ. Prefix mandatory nodes with **; organizer nodes with !!.
Primary Question	The main question. Leave blank to add an alternate answer to the previous question.
Alternate Question	Optional. Add one per row for multiple alternates.
Tags	Tags for the question or alternate question.
Answer	The response. Supports plain text, scripts (SYS_SCRIPT; <js>), channel-specific (SYS_CHANNEL_<name> SYS_TEXT; <text>), or dialog trigger (SYS_INTENT; <dialog ref id>).
Extended Answer-1 / Extended Answer-2	Optional. Use when the response is lengthy.
ReferenceId	Optional. External source reference.
Display Name	Optional. Shown to users in ambiguity scenarios.

Nodes Section

Column	Description
Node	Section header; leave blank in data rows.
Nodepath	Path to the node/tag.
Tag	Required for tag settings; leave blank for node.
Precondition	Conditions required to qualify this node/tag.
outputcontext	Context populated by this node/tag.
Traits	Traits for this node/tag.

Synonyms Section

Column	Description
Synonyms	Section header; leave blank in data rows.
Phrase	The term or phrase for which to add synonyms.
Synonyms	Comma-separated synonym values.

To enable Bot Synonyms in KG term identification:

• Use confidenceConfigs as the header.
• Set parameter to useBotSynonyms and value to true or false.

KG Params Section

Column	Description
KG Params	Section header; leave blank in data rows.
lang	App language code (for example, en for English).
stopwords	Comma-separated stop words.

Traits Section

Column	Description
Traits	Section header; leave blank in data rows.
lang	App language code.
GroupName	Trait group name.
matchStrategy	Pattern or probability (ML-based).
scoreThreshold	Threshold (0-1) when matchStrategy is ML-based.
TraitName	Name of the trait.
Training data	Utterances for the trait.

For Taxonomy-based KG (FAQs linked to other FAQs):

Field	Description
faqLinkedTo	Que ID of the source FAQ this FAQ is linked to.
faqLinkedBy	Que IDs of FAQs linked to this FAQ.
isSoftDeleted	Identifies FAQs that are deleted but still have linked FAQs.

JSON File

Download the Sample

1. Go to Automation AI > Knowledge AI > FAQs > ⋯ (more) > Import.
2. Back up when prompted, then click Proceed.
3. Click Sample JSON to download.

JSON Property Reference

Property	Type	Description
FAQ	Array	Contains question, answer, terms (leaf to first-level node), and alternate questions.
question	String	Primary question (inside FAQ array).
answer	String	AI Agent response (inside FAQ array).
terms	Array	Leaf node and parent nodes up to the first-level node.
refId	String	Optional external source reference.
Alternate Questions	Array	Alternate questions and their terms (leaf to first-level node).
Synonyms	Object	Arrays of terms and their synonyms.
Unmappedpath	Array	Nodes with no questions, and their parents up to first-level node.
Traits	Object	Trait names as keys; arrays of utterances as values.

For Taxonomy-based KG:

Field	Description
faqLinkedTo	Identifies the source FAQ.
faqLinkedBy	Lists the FAQs linked to this FAQ.
isSoftDeleted	Identifies deleted FAQs that still have linked FAQs.

---