Administrative Inquiries and Responses for Territorial Spatial Planning Formulation and Implementation Evaluation: Knowledge Base Retrieval and Recall

Administrative Inquiries and Responses for Territorial Spatial Planning Formulation and Implementation Evaluation: Knowledge Base Retrieval and Recall

What the data for this category looks like

The daily documents for this category include territorial spatial overall plans, detailed plans, special plans, planning implementation evaluation reports, and public opinion summaries. Data sources include the territorial spatial basic information platform, natural resource survey and monitoring data, population and economic statistics, and remote sensing images. The update frequency is annual. Document structures include spatial coordinate fields, planning constraint indicators, regional scope parameters, and text content from public feedback. Some documents include structured spatial layer metadata.

The annual update nature of the data for this category requires the knowledge base synchronization cycle to match the annual update rhythm, to avoid frequent full synchronization that consumes computing resources. The multi-source and mixed data types, including structured spatial metadata, professional text reports, and colloquial public opinions, require retrieval and recall to support both text semantic matching and structured field association. Documents contain clear regional scope parameters, so retrieved results must be associated with the spatial constraints of the target planning area, to avoid returning invalid cross-regional content. The text in public opinion summaries uses non-standard expressions, so the semantic matching threshold for retrieval needs to be adjusted to adapt to the semantic recognition accuracy of this type of content.

How to set configurations

Reference only; verify against the target FastGPT release and deployment documentation.

Configuration ItemRecommended ValueRationale
similarity0.72–0.80Adapts to the colloquial public opinion text in this category, avoids recalling irrelevant content if the value is too low, or missing valid opinions if the value is too high
limitTop 10 entriesBalances the coverage of retrieval results and response speed, adapts to the conventional need for multi-source reference during planning formulation
searchModeFull-text retrievalAdapts to the multi-source mixed document types of this category, supports joint matching of text and structured fields
usingReRankEnabledImproves the sorting accuracy of content related to planning constraint indicators, avoids placing semantically similar but irrelevant results at the top
datasetSearchUsingExtensionQueryEnabledAutomatically expands query terms based on professional terminology for planning formulation, adapts to retrieval needs for industry-specific nouns
collectionIdsKnowledge base collections bound to planning regionsAccurately limits the retrieval scope, avoids recalling invalid cross-regional documents

Three common errors

  • Only the file collection ID of a single knowledge base is returned when calling the API for querying, without covering all bound knowledge bases. This is caused by incorrect configuration of the collectionIds parameter, failing to associate the knowledge base collections of both target planning regions at the same time.
  • The retrieval request returns an error message containing "DDG detected an anomaly in the request". This is caused by mistakenly enabling external network search configuration, without limiting retrieval to only internal knowledge bases.
  • Garbled characters appear after uploading CSV data exported from WPS to the knowledge base. This is caused by failing to convert the CSV file encoding to UTF-8 in advance, leading to character encoding conflicts during knowledge base parsing.

How to confirm the configuration is complete

  • View the collectionIds parameter bound to the knowledge base, confirm that it includes all knowledge base collections associated with the target planning regions.
  • Run a test query, check whether the semantic matching scores of the returned results fall within the preset 0.72–0.80 range.
  • After enabling the re-ranking function, verify whether the top-ranked results include core business content such as planning constraint indicators and regional scope.
  • Call the API interface, confirm that the returned results include the file collection ID field and cover all bound knowledge base content.

The values provided here 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.