Search and classify free-text data by “concept” inside your existing PostgreSQL database.
Without introducing a new vector database or graph database, Concept Index enables you to extract semantically similar groups from VoC data, inquiry histories, free-text survey responses, sales notes, project summaries, and other business records.
Concept Index is scheduled for release in August 2026. The initial supported database is PostgreSQL.
Free-text data is difficult to use, even when it is already stored in an RDB.
Companies accumulate large amounts of valuable text data, such as customer feedback, inquiry histories, sales notes, and survey responses. However, conventional RDBs make it difficult to handle “semantic similarity” that cannot be captured by keyword search or fixed categories.
Concept Index assigns “concept nodes” to each record.
Concept Index analyzes business data that includes text columns and organizes records with similar meanings into concept nodes. As a result, each record gains information not only as an ID or category, but also as a member of a specific concept group.
Designed for these types of business data.
Concept Index is suitable when you want to extract groups of data that are semantically similar but difficult to find through keywords alone.
VoC Analysis
Extract customer groups with similar complaints, requests, or expectations, and use them for product and service improvement.
Inquiry Histories
Identify similar consultation topics and recurring trouble patterns to improve response quality.
Free-text Survey Responses
Organize respondents’ interests, concerns, and expectations as meaningful semantic groups.
Sales Notes
Extract similar sales patterns and deal types to support the knowledgeization of sales activities.
Project Summaries
Refer to past similar projects by concept and use them as material for proposals and decision-making.
Business Data Exploration
Discover latent patterns in operational data that are difficult to see through fixed categories.
This is not RAG. Concept Index is designed to find similar data groups, not individual answers.
Concept Index is not a search engine for returning pinpoint documents in response to a question. Its purpose is to extract conceptually similar groups and segments from records that include free-text data. If you want to build a question-answering system or knowledge base, please use ThinkNavi Knowledge Base Builder instead.
From Top-K search to concept-based segment extraction.
Concept Index is not designed to search for the nearest document one by one. Instead, it uses pre-built concept nodes and neighborhood relationships to handle business-meaningful groups of data.
| Comparison Item | General Vector Search | Concept Index |
|---|---|---|
| Main Purpose | Find documents close to a query | Extract semantically similar data groups |
| Data Infrastructure | Often requires an additional vector database or dedicated infrastructure | Uses existing RDB infrastructure |
| Search Unit | Individual documents or individual records | Concept nodes, neighboring nodes, and segments |
| Suitable Uses | RAG, FAQ, document search | VoC analysis, survey classification, project group extraction |
Connector-based delivery for easy integration into existing environments.
Concept Index is introduced by adding a concept index to the existing RDB where your business data is stored. Rather than building a large new AI infrastructure, it extends the way you can use your current database.
Windows / Linux Support
Provided as an executable file to reduce the burden of setting up a Python environment.
Set API Key and DB Connection Information
Specify the target table or View and run the concept index construction process.
Write Results Back to the RDB
Make concept node IDs, neighborhood relationships, and classification results available inside the RDB.
In the future, extract concept-based segments using natural language.
As a future extension, we are considering Concept Query, a feature that allows users to specify complex segment conditions in natural language.
Customers with high profitability but also high churn risk
Concept-based customer extraction that combines quantitative data with free-text data.
Groups with low price sensitivity and strong premium orientation
Discovery of purchasing tendencies that are difficult to see through simple attribute classification.
Segments with many inquiries but also high satisfaction
Understand not only problem frequency, but also improvement potential and relationship strength.
※ Concept Query is currently under consideration as a future extension.
Limited Pre-order Offer for Concept Index
ConceptMiner Concept Index, scheduled for release in August 2026, is available at a limited first-year pre-order price.
Standard Annual License
- Supported DB: PostgreSQL
- Basic Concept Index functionality
- Concept classification of free-text data
- Writing concept node information back to the RDB
Pre-order Price
- Scheduled for release in August 2026
- Pre-order price applies to the first year only
- PostgreSQL-compatible version
- Use-case confirmation before introduction
Add a new concept-based search capability to your existing RDB.
VoC data, inquiry histories, free-text survey responses, sales notes, project summaries — use the text data already accumulated inside your company as meaningful semantic groups.