Citation Source and Traceability for Iron Ore Investment Research Knowledge Base Construction

Iron ore investment research data includes three core source types: spot, futures, and industry reports.

What the data for this category looks like

Iron ore investment research data includes three core source types: spot, futures, and industry reports. Spot price data comes from port spot trading platforms and industry news sites. Futures data is sourced from the Dalian Commodity Exchange. Supply and demand and freight rate data comes from regular reports published by steel industry associations and international mining institutions. Update frequencies vary: spot data updates daily, futures data synchronizes within one hour after market close, and industry reports release weekly or monthly. Document structures primarily use structured tables, with core fields including ore grade, origin, delivery grade, unit price, and more. Common units are yuan/wet ton, US dollars/dry ton, and yuan/ton·kilometer.

Constraints on citation source and traceability workflows

Multiple data sources use inconsistent field naming. For example, English reports from international mining institutions use "Fe content", which corresponds to "grade" in domestic industry documents. Traceability workflows require configured field mapping rules. Without these rules, cross-language data association and matching cannot be completed. Update frequencies differ widely across data sources. Incremental update tasks must adapt to multiple update cycles. Otherwise, data lag or duplicate indexing will occur. Some industry associations and third-party platforms implement anti-crawling mechanisms. Access to traceability links requires handling verification logic. Without this logic, invalid traceability results will be generated. The field correspondence between English literature and Chinese data is complex. Without configured matching rules, Chinese queries cannot recall content from English data sources.

Configuration Settings

Configuration ItemRecommended ValueRationale
similarity_threshold0.78–0.82Iron ore data has unified core identification fields. An overly high threshold will miss relevant data sources, while an overly low threshold will include non-target category data
recall_top_kTop 6 entriesA single iron ore report contains multiple sets of data. Too many recalled entries will increase traceability organization costs, while too few will miss critical data sources
incremental_update_interval2 times per daySpot data updates daily, and futures data updates after market close. Two crawls per day cover the update cycles of both data source types
reference_match_fieldsGrade, origin, delivery gradeCore identification fields for iron ore data, used to accurately match traceability results with query keywords
link_validation_enabledEnabledSome industry associations and third-party platforms have anti-crawling mechanisms. Validation filters invalid or expired traceability links
retry_times_on_timeout3 timesSome data source websites load slowly. Retries improve crawl success rates and avoid traceability task interruptions

The parameter values provided on this page are common starting points for configuration. Actual values are affected by material form, data volume, and business rules. Specific issues require case-by-case analysis. It is recommended to test against your own samples before finalizing settings.

Three common misconfigurations

  • Phenomenon: After submitting a Chinese query containing "62% iron ore fines", no English Platts Energy reports from the knowledge base are recalled. Cause: No Chinese-English field mapping rules are configured for reference_match_fields, so Chinese and English data cannot be associated and matched.
  • Phenomenon: Incremental update task logs show 403 Forbidden errors. Cause: link_validation_enabled is not enabled, and no anti-crawling verification rules are configured. This results in blocked access to links from some industry association websites.
  • Phenomenon: Irrelevant data for both iron ore fines and lump ore appears in traceability results. Cause: The similarity_threshold is set below 0.75, which recalls non-target category data that does not meet similarity requirements.

How to confirm successful configuration

  • Submit a query containing "Qingdao Port 62% iron ore fines spot price", then check if the traceability column of the returned results marks the corresponding port spot trading platform link or document name.
  • View the running logs of incremental update tasks, and confirm that there are no 403 or timeout error records in the past 7 days.
  • Import English Platts Energy reports and Chinese industry reports, then submit a query containing the corresponding grade. Check if traceability results match both types of documents.
  • Upload an iron ore supply and demand report in Notion format, then check if the parsed document is included in the traceability index and can be correctly associated with query results.

Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-14.