What Data for This Category Looks Like
Data for this category primarily comes from the real estate registration information system, cadastral survey database, housing transaction online signing system, tax system, and electronic certificate repository. The update frequency is daily. Documents processed daily include registration application forms, ownership certificates, surveying and mapping results, tax payment vouchers, real estate ownership certificates, and real estate registers. Data exists in a mixed form of structured fields and unstructured documents. Fields include real estate location, right holder information, ownership source, restriction status, surveying and mapping area, and others. Field formats vary by document type.
Constraints Imposed by These Characteristics on Knowledge Base Retrieval and Recall
Cross-system, multi-source data requires retrieval to cover both structured fields and unstructured documents, and match both in-system structured data and uploaded various documents. Daily update frequency requires the retrieval pipeline to support real-time or near-real-time data source synchronization, to avoid retrieving expired data. Large differences in document field formats require adapting field extraction rules for different documents, to avoid recall result deviations due to format mismatches. Ownership source and restriction status are core verification fields. Retrieval must prioritize highly relevant results containing these fields. Otherwise, subsequent registration errors or property rights disputes may occur, and time spent on subsequent verification links will increase.
Configuration Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Real estate registration data requires accurate matching of core fields such as ownership and location. This range filters low-relevance results and avoids incorrect association of ownership information |
limit | Top 5–8 results | Ledger summary needs to cover major ownership sources and restriction statuses. This number balances recall completeness and retrieval efficiency |
searchMode | hybrid | Retrieval of both structured fields and unstructured documents is required. Hybrid retrieval balances recall performance for both types of data |
usingReRank | Enabled | Real estate data has strong field relevance. Reranking can improve the sorting priority of results matching core ownership fields |
datasetSearchUsingExtensionQuery | Enabled | Extended queries can supplement synonymous expressions for professional terms such as ownership and location, improving recall matching accuracy for cross-system data |
collectionIds | Exclusive collection IDs divided by business domain | Real estate registration data must be isolated from data of other categories. Exclusive collections avoid interference from cross-category retrieval |
Three Common Configuration Errors
- Search test returns no results. The cause is failure to configure exclusive business domain
collectionIds, or uploaded registration documents have not completed parsing and import. - Retrieved reference count does not meet preset requirements. The cause is incorrect use of
similarityto control returned result count, orlimitparameter value does not match the ledger summary count requirement. - Hybrid retrieval time exceeds expectations. The cause is failure to restrict retrieval scope, or enabling
usingReRankwithout adjusting the reranking recall upper limit.
How to Verify Correct Configuration
- Run a search test, enter a query containing real estate location and right holder information, check whether returned results include documents and structured data from the corresponding business domain.
- Adjust the
limitparameter, verify whether the number of returned results matches the preset value. - After enabling
usingReRank, compare result sorting before and after reranking, confirm that the priority of results with core ownership fields has been improved. - Check document parsing status, confirm that uploaded registration application forms, ownership certificates and other documents all show parsing completed.
The configuration values listed above are common starting points, and should be evaluated against the reader’s own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-28.