...
Below are the steps to create an Artifact using the FAQ Template:
On the Create Artifact window, select Template.
On the Create Artifact window, navigate to the Using Template tab.
Select FAQ Template as Template Type.
Add Artifact Name.
Add Summary. For the Metadata-based search,It is important that the summary is well-formed and describes the information in the document well. The text in this field is used by the NLP engine to generate Metadata.
Click on Add Q&A Pairs.
Add Q&A Pairs
In this step, the system allows you to add one or more Q&A pairs. Information in these pairs is used to generate related metadata attributes like Parent Topic Action, and Motivation for each Q&A pair.
Click on Add Q&A Pair to add new QnA pairs
On Add Questions & Answer pair screen, add details. Once done, click on the Add Q&A Pair button.
The Q&A is added to the list. You may add multiple Q&A pairs to the Artifact.
On the Artifact page, Click on Generate Ontology Create.
The Artifact is now created and available in the Knowledge Store. Information added to these Q&As should be well-formed as this ensures that the Ontology process is able to extract correct values for metadata attributes.
Based on the Artifact Publishing Mode configuration for your tenant and the identified metadata, the Artifact state is determined. For more information, refer to Tenant Configurations.
Generate Ontology
...
| Info |
|---|
System prevents creation of duplicate FAQs in an artifact . Using in-built Text similarity NLP service, the system identifies and ignore duplicate FAQs with Text similarity of over 90% against the existing FAQs in the artifact |
Processing the Artifact
For Metadata Search:
Once the Artifact is added, Luma generates ontology for the artifact. The Metadata is added to the artifact, which is then used to identify Knowledge during the search. When Open AI features system performs the following:
Summarize Artifact- If Automatic Summarization is enabled for your tenant, the content in the artifact is used to generate a Summary. In case the curator adds the Summary during artifact creation, the step is skipped.
Generate Keywords and Key Phrases: Based on the Summary and FAQs added, metadata attributes like Topic, Subject Action, Motivation, and Key Phrases are generated. If Automatic Keyword and Key Phrase generation is enabled for your tenant, the system automatically generates valid and meaningful full keywords and key phrases. On click of Generate Ontology, the NLP engine parses the information added to the Artifact and identifies the metadata, which is used to understand the user’s intent during the search.
As a Curator, you should verify and ensure that metadata generated by the system is correct and Artifact is related to the correct Topic. In case any changes are required,
Click on Edit.
Add the required keyword to metadata fields and Save Changes.
Click on View Artifact to view the Artifact in the Knowledge Store. You can Publish the Artifact making it available for end-users to search and consume. Refer to Knowledge Store for more information.
Click on Go To Dashboard, to navigate to the Luma Knowledge dashboard
| Info |
|---|
|
For Semantic Search:
Once the Artifact is added, No Curation of the artifact is required. Ontology, Summary, and FAQ generation are skipped for Semantic Search. The system automatically does the following to generate the searchable artifact content.
Data Extraction and Chunking: In this phase, text is collected or extracted from the source document. The extracted text is then divided into manageable chunks, typically at the semantic level (e.g., paragraphs).
Vectorization: The text chunks undergo embedding, where they are converted into vectors or numerical forms. This conversion is crucial for enabling efficient and accurate search capabilities. The resulting embeddings are stored in a vector database, ready for search and retrieval operations.
| Info |
|---|
|