What Data for This Category Looks Like
Financial report-related data for the glass industry category primarily comes from industry association monthly production bulletins, listed company periodic reports, and public bidding data. Data update cycles fall into three categories: monthly, quarterly, and annual. Monthly data includes ex-factory average prices and capacity utilization rates. Quarterly and annual data includes core metrics such as revenue structure, unit production costs, and gross profit margins. Most documents combine structured tables and written analysis, with fields including product type, production volume, unit price, revenue composition, energy consumption indicators, and others. Common units include tons, yuan per square meter, ten thousand yuan, kilograms of standard coal per weight box, and similar units.
What Constraints Do These Characteristics Impose on the Knowledge Base Retrieval and Recall Link
Glass industry financial report data mixes high-frequency monthly indicators and low-frequency annual financial reports, so the retrieval link must support precise filtering by update time and data type. There are many structured fields with inconsistent units, so automatic unit alignment must be performed before recall to avoid invalid recall caused by unit differences. Long documents contain data across multiple product categories, so product type and time tags must be retained during segment processing to ensure retrieval results match content for corresponding product segments. Industry data and corporate financial reports are stored together, so data source tags must be used to distinguish entries and prevent cross-entity data confusion.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
PARSE_FILE_TIMEOUT_SECONDS | 300 seconds | Glass financial report documents contain long tables and multi-page graphics and text, with longer parsing times than general documents |
Segment Length | 800–1200 characters | Glass financial reports have dense structured fields, so segments must retain the integrity of field associations to avoid disrupting data logic during splitting |
Recall Count | Top 8 results | Glass financial report analysis requires coverage of multi-dimensional indicators such as production volume, prices, and costs, so a sufficient recall volume is needed to support matching |
Similarity Threshold | 0.72–0.80 | Field matching for structured data requires balancing precision, to avoid recalling cross-category or irrelevant glass industry data |
UPLOAD_FILE_MAX_SIZE | 2000 MB | Annual industry compilations and corporate financial report documents have large file sizes, so upload limits must be relaxed |
Reranked Return Count | Top 4 results | Prioritize returning the most relevant structured data to improve retrieval efficiency for financial report analysis |
The parameter values provided on this page are common starting points for configuration. Actual values are affected by material format, data volume, and business rules. Specific issues require specific analysis, and it is recommended to test on your own samples before finalizing settings.
Three Common Mistakes
- Issue: After a glass financial report document with embedded images is uploaded to the knowledge base, images cannot be displayed normally in retrieval results. Cause: The default parsing configuration does not enable image indexing and storage, only text content is extracted, and image links are not included in the retrieval library.
- Issue: After HTML interface documents generated by Javadoc are uploaded, the knowledge base fails to parse structured interface fields and parameters. Cause: The default document parser only supports plain text and standard table formats, and dynamically rendered content embedded in HTML cannot be directly recognized.
- Issue: When retrieving specific capacity indicators for the glass industry category, the number of recall results is much lower than expected. Cause: The recall count configuration is set too low, failing to cover enough relevant document fragments, resulting in some granular data not being recalled.
How to Confirm Proper Configuration
- A small glass monthly financial report document is uploaded. The segmented content parsed by the knowledge base is reviewed to confirm the segment length matches the preset configuration.
- A retrieval request containing specific indicators for the glass industry category is submitted. The number and relevance of recall results are verified to confirm they match business requirements.
- A large annual industry compilation document is uploaded. Confirm that no file size limit related errors are triggered during the upload process.
- Glass financial report documents containing embedded images are retrieved. Confirm that retrieval results include associated image information if image indexing configuration has been enabled.
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.