Data Characteristics for This Category
Documentation for culture media and consumables primarily includes product manuals, technical specifications, Safety Data Sheets (SDS), and batch analysis reports. Suppliers typically provide these documents. Updates are infrequent, usually occurring only when product formulations, specifications, or regulatory requirements change. Product manuals often contain sections like product name, model, components, storage conditions, usage instructions, and quality control indicators. Technical specifications focus more on detailed physical and chemical indicators and performance parameters. SDS documents detail hazard information for components. Common fields include chemical component percentages, pH values, osmolality, viscosity, particle size, expiration dates, and batch numbers. Units vary and include g/L, mg/L, pH units, mOsm/kg, cP, and μm.
Constraints Imposed by These Characteristics on "Document Parsing and Chunking"
The specific structure and fields of culture media and consumable documents impose particular requirements on document parsing and chunking. First, tabular data (e.g., component lists, quality control standards) in product manuals and technical specifications are core information. Parsing must preserve table structures and correctly associate fields with values. Second, critical safety data, such as hazard information and first aid measures in SDS documents, are often presented in specific sections or formats. These critical pieces of information must remain intact during chunking to avoid truncation. Third, due to the variety of units, the parser must accurately identify and retain unit information to prevent semantic errors caused by separating units from numerical values. Finally, the low frequency of product updates means that initial knowledge base construction requires significant effort to ensure accurate document parsing. Subsequent maintenance pressure is relatively low, but new products or version updates necessitate precise re-parsing and chunking.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
Chunk Length | 500–800 characters | Paragraphs in culture media and consumable documents often contain many technical terms and parameters. Increasing chunk length helps maintain contextual completeness and avoids splitting critical information. |
Overlap Length | 50–100 characters | Ensures sufficient overlap between adjacent chunks to handle critical information spanning paragraphs, such as continuing component lists or sequential operating procedures. |
Parsing Strategy | Table First | Product manuals and technical specifications contain extensive tabular data. This strategy ensures correct identification and extraction of table content, preserving its structured information. |
Max Paragraph Depth | 3 | Most culture media and consumable documents do not have excessively deep hierarchical structures. This depth is sufficient to cover most sections and subsections. |
File Types | PDF, DOCX, XLSX | Suppliers typically provide documents in these formats, ensuring the parser covers mainstream file types. |
Three Common Mistakes
- After uploading tabular data to the knowledge base, some row data is lost because the parser incorrectly identifies table boundaries or continuity of tables spanning multiple pages, leading to incomplete data.
- After uploading PDF documents containing images, text content within images is not extracted because the default parsing model lacks image OCR capabilities.
- When querying product parameters, results show numerical values separated from units or incorrect units because the parser separates values and units during chunking or fails to accurately identify unit symbols.
How to Verify Correct Configuration
- Upload typical product manuals and technical specifications. Examine the parsed chunks to ensure table structures, key parameters, and unit information are complete and accurate.
- Upload PDF documents containing complex charts or scanned images. Verify whether the parsed results include text information around or within images to assess if OCR capability meets expectations.
- Perform query tests on the knowledge base. Ask questions about key information such as product components, storage conditions, and safety data. Evaluate the accuracy and completeness of retrieved results.
- Check parsing logs for any records of file parsing failures or timeouts. Adjust relevant configuration parameters based on error messages.
The values provided are common starting points and should be measured against your 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 2026-09-21.