I am trying to get chatgpt 3.5 to refine a dataset I have extracted from PDFs. The dataset will act as a knowledge base for an RAG system. The initial tests show great improvement, except for the bot returning the answer with a comment at the end such as "here you have the text with formatted tables, rewritten sentences, etc... ". Of course, I don't want to have this in my knowledge base, as I want to apply this to a lot of text. How can I avoid this? Note: look at line 10
# Prompt
prompt_text = """As an assistant, your objective is to improve text and table readability. Here's your guide:
1. Reframe sentences and sections for better understanding.
2. Eliminate unclear text. Example: "Text containing excessive symbols or gibberish."
3. Shorten text where possible without losing information. Suggestion: Summarize lengthy phrases when feasible.
4. Rectify poorly formatted tables. Example: Adjust column alignment for clarity.
5. Preserve clear, understandable text as it is. Example: "Use direct and easily comprehensible sentences."
6. If text is entirely unclear or ambiguous, refrain from providing a response. Example: "Incomprehensible or garbled content."
7. Remove standalone numbers or letters not associated with text. Example: "Eliminate isolated digits or letters lacking context."
8. Exclude :selection marks, x , and other non-factual elements.
9. Ensure modifications maintain the original text's clarity and don't compromise conveyed information.
10. Return the revised text without any additional comments before or after.
Please revise the following text based on the listed guidelines: {element} """
prompt = ChatPromptTemplate.from_template(prompt_text)