Relevance Tuning

This feature allows users to increase the response of a result when a query contains terms that match the attribute.

  • We recommend connecting to Watson Discovery, watsonx Discovery, or Elastic AppSearch to utilize this feature.

Dynamic Filter Query

This feature allows for users to apply filters to their queries based on specific criteria in order to refine their search results.

  • We recommend connecting to Watson Discovery or watsonx Discovery to utilize this feature.

This feature utilizes numerical representations of data, known as vectors, to conduct searches and identify relevance. In traditional leucine searches, documents are indexed based on keywords and queries are matched to documents containing those exact keywords. Vector searching utilizes semantic relationships to find related objects in the documentation that share similarity. This approach is ideal for broad or fuzzy queries, and improves the depth and breadth of searching and querying different types of data.

  • We recommend connecting to ElasticSearch for document-oriented vector search.
  • We recommend connecting to Milvus or Pinecone for flexible, and scalable data handling with high-performance vector search.
  • Additonally, we recommend Amazon Kendra or Amazon Bedrock for managed vector search to aid in data chunking, embeddings, and indexing algorithm choices.

External Embedding Model Support

This feature utilizes an external embedding model to create vector embedding for indexing content. Upon query, the embedding model creates embeddings for that query, and uses them to query the database for similar vector embeddings for answer generation.

  • We recommend connecting to Pinecone or Milvus to utilize this feature.

KnowledgeBase Capabilities

Features Chart
KnowledgeBase Supported Search Types Query Filters Document Prioritization (Re-Sort) Relevance Tuning Dynamic Filter Querying Full Document Retrieval External Embedding Model Support
Watson Discovery Lucene
watsonx Discovery Lucene, Vector, Hybrid
Elastic AppSearch Lucene
ElasticSearch Lucene, Vector, Hybrid
Amazon Kendra Vector (Managed)
Amazon Bedrock Vector (Managed)
OpenSearch Lucene
Pinecone Vector
Milvus Vector

How to Configure NeuralSeek with a KnowledgeBase

Setting up the integration to KnowledgeBase is done in Configure > Corporate Knowledge Base Details page.

See the Configure reference page for more details.

Ⓒ 2024 NeuralSeek, all rights reserved.