表格数据提取验证器

从PDF或图片中的表格提取数据时使用,自带验证和置信度评估,避免提取错误。

提示词

You are extracting data from a table (image or text).
Task: Extract table data into a structured format.
Instructions:
1. Identify column headers (or describe columns if no headers)
2. Extract each row of data
3. Validate data types in each column:
   - Numbers: Note if they appear to be currency, percentages, counts
   - Dates: Note the format used
   - Text: Note if they appear to be codes, names, or descriptions
4. Flag any cells that are unclear or potentially misread

Output Format:
Table Structure:
- Columns: [number]
- Rows: [number, excluding header]
- Data Types: [column → type mapping]
Extracted Data:
[Reproduce table in markdown format]
Validation Notes:
- Uncertain cells: [list with row/column reference]
- Format inconsistencies: [list any issues]
- Missing data: [list empty cells]
Confidence: [HIGH/MEDIUM/LOW] based on clarity

Constraints:
- Mark uncertain values with [?]
- Do not fill in missing values
- Preserve original formatting of numbers/dates

示例输出

输出:表格结构信息、提取后的Markdown表格、验证说明(不确定单元格+格式不一致+缺失数据)、置信度评级(高/中/低)。不确定值用[?]标记。

0

评论 0

更多

登录后可点赞、收藏、评论和举报。

还没有评论,先发起一个具体问题。