diff --git a/fieldExtraction/src/prompts/investment_prompts.py b/fieldExtraction/src/prompts/investment_prompts.py index 2a1c4c2..c0d8e7c 100644 --- a/fieldExtraction/src/prompts/investment_prompts.py +++ b/fieldExtraction/src/prompts/investment_prompts.py @@ -259,18 +259,28 @@ class FieldSet: def EXHIBIT_HEADER(context): return f"""Analyze the following contract excerpt and extract the full header found. -Include ALL text that appears to be part of the exhibit/attachment header. Include all of the following when present: -* Full Attachment names and numbers (e.g., "Attachment C: Commercial-Exchange") -* Complete Exhibit numbers/letters with full titles (e.g., "Exhibit 1 - Medicare") -* Any other relevant subtitles and identifiers, including but not limited to 'Article', 'Amendment', 'Schedule', 'Addendum', 'Appendix', or similar. +Capture the complete hierarchy of document identifiers present in the text. The following keywords and phrases should be considered as exhibit/attachment headers: +* EXHIBIT +* ATTACHMENT +* ARTICLE +* AMENDMENT +* SCHEDULE +* ADDENDUM +* APPENDIX + +Include the following when present: +* Full header names and numbers/letters (e.g., "Attachment C: Commercial-Exchange") +* Any other relevant subtitles and identifiers * Any words in the header that may specify what type of information is contained, such as "Compensation Schedule" or "Definitions". * Provider/Entity names when listed with exhibit information -Here are some very important rules: -* Do NOT abbreviate or summarize any part of the exhibit names -* Do NOT include page numbers -* Do NOT include any Section language from your extraction of the header, even if it appears to be a part of the header. Sections are not considered part of the header. -* Do NOT include docusign IDs and other metadata. These do not constitute a header. +- Rules for inclusion: + * Do NOT abbreviate or summarize any part of the exhibit names + * Do NOT include page numbers + * Do NOT cherry-pick only certain parts - capture the complete hierarchy + * Do NOT include docusign IDs and other metadata + * Include ALL text that appears to be part of the header + * Keep exact capitalization and formatting as shown in document If there is no Exhibit Header present, return |N/A|. Do not fill in values from the examples above, these are only examples. @@ -279,7 +289,7 @@ Here is the text to analyze: Before returning any output, ensure that all instructions for inclusion were followed. -Briefly explain your answer, then enclose your final answer in |pipes|. +Enclose your final answer in |pipes|, followed by a brief explanation. """ def EXHIBIT_LINKAGE(header1, header2):