引導式測驗 · 中等 等級
AI產品屬性擷取測驗
從文字和圖像中提取目錄字段,同時保留模式、單位、來源證據和人工檢查不確定屬性。
問題 1 的 8
目錄團隊在提取產品屬性之前應該定義什麼?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- 目錄團隊在提取產品屬性之前應該定義什麼?
- A category-specific schema with types, units and allowed values
- A prompt asking for any details the model can imagine
- One universal field list for every industry
- A rule to publish all high-confidence outputs
- 產品形象靠自身無法可靠地建立什麼?
- That a visible zipper appears in the photograph
- An item’s hidden fabric composition when no source states it
- That the image contains a blue object
- Whether the package shows a printed label
- 為什麼在規範化屬性時保留來源值?
- To make duplicate products appear unique
- To avoid validating the normalized field
- To trace conversions and review how the value was represented originally
- To remove supplier information
- 當兩個供應商文件在材料方面有分歧時,管道該怎麼做?
- Choose whichever value has more words
- Average the two material names
- Publish both values without context
- Flag the conflict for review instead of silently overwriting
- 為什麼將尺寸或重量視為帶有單位的鍵入欄位?
- A value without its unit can be misread or converted incorrectly
- Units are only needed for images
- The model can infer the retailer’s preferred unit every time
- Typed fields prevent supplier errors automatically
- What does a confidence score represent?
- Proof the attribute is factually correct
- A model estimate that should be checked against evidence
- An official product certification
- The supplier’s approval
- Which validation is useful for extracted apparel size?
- Check whether the photo is square
- Ensure the brand name is capitalized
- Map the value to the retailer’s allowed size vocabulary and variant
- Replace every regional size with a single value
- How should teams evaluate an attribute extractor?
- Use one aggregate score for the entire catalog
- Check only the highest-confidence examples
- Count how many fields the model outputs
- Measure accuracy and corrections by field, category and supplier