Openai utils

embedding_content(prompt, model_name='text-embedding-3-small')

Parameters:

Name	Type	Description	Default
prompt		The prompt to generate the embedding for	required
model_name		The model to use for generating the embedding	'text-embedding-3-small'

Source code in TimelineKGQA/openai_utils.py

def embedding_content(prompt, model_name="text-embedding-3-small"):
    """
    Args:
        prompt: The prompt to generate the embedding for
        model_name: The model to use for generating the embedding

    """
    response = client.embeddings.create(input=prompt, model=model_name)

    return response.data[0].embedding

paraphrase_medium_question(question, model_name='gpt-4o')

Paraphrases the given question using the OpenAI model specified.

Parameters:

Name	Type	Description	Default
question	str	The question to paraphrase.	required
model_name	str	The model to use for paraphrasing.	'gpt-4o'

Returns:

Name	Type	Description
str	str	The paraphrased question.

Source code in TimelineKGQA/openai_utils.py

def paraphrase_medium_question(
    question: str,
    model_name: str = "gpt-4o",
) -> str:
    """
    Paraphrases the given question using the OpenAI model specified.

    Args:
        question (str): The question to paraphrase.

        model_name (str): The model to use for paraphrasing.

    Returns:
        str: The paraphrased question.
    """
    prompt_text = f"Paraphrase the following question: '{question}'"
    try:
        # Some examples include:
        # Who is affiliated with the organization during a given time.
        # Which or what's the organization's name a specific guy is affiliated to.
        # When/During/when is start time ...
        # Etc.
        # If there is a statement from beginning of time to the end of time, this will mean it is always true for the whole timeline.
        response = client.chat.completions.create(
            model=model_name,
            messages=[
                {
                    "role": "system",
                    "content": """You are an expert on paraphrasing questions.
                                  ??? is the masked out answer
                                  You job is paraphrasing this into a natural language question.
                                  The missing part can be someone, some organisation or some time.
                                  Representive question types include:
                                  - When/Before/After/During/(temporal conditions calculated based on Timeline Operation) event A happen, who is leader of organization B?
                                  - Is event A before/after/during event B?
                                  And we should not mention any specific time in the question, it is a type of implicit temporal questions.
                                  """,
                },
                {
                    "role": "user",
                    "content": prompt_text,
                },
            ],
            max_tokens=100,
            temperature=0.8,
            stop=["\n"],
        )
        paraphrased_question = response.choices[0].message.content
        return paraphrased_question
    except Exception as e:
        print(f"An error occurred: {e}")
        return ""

paraphrase_simple_question(question, answer=None, answer_type=None, model_name='gpt-4o')

Paraphrases the given question using the OpenAI model specified.

Parameters:

Name	Type	Description	Default
question	str	The question to paraphrase.	required
answer	str	The answer to the question, which can help in generating a context-aware paraphrase.	None
answer_type	str	The type of the answer, which can help in generating a context-aware paraphrase.	None
model_name	str	The model to use for paraphrasing.	'gpt-4o'

Returns:

Name	Type	Description
str	str	The paraphrased question.

Source code in TimelineKGQA/openai_utils.py

def paraphrase_simple_question(
    question: str,
    answer: str = None,
    answer_type: str = None,
    model_name: str = "gpt-4o",
) -> str:
    """
    Paraphrases the given question using the OpenAI model specified.

    Args:
        question (str): The question to paraphrase.
        answer (str, optional): The answer to the question, which can help in generating a context-aware paraphrase.
        answer_type (str, optional): The type of the answer, which can help in generating a context-aware paraphrase.
        model_name (str): The model to use for paraphrasing.

    Returns:
        str: The paraphrased question.
    """
    prompt_text = f"Paraphrase the following question: '{question}'"
    try:
        # Some examples include:
        # Who is affiliated with the organization during a given time.
        # Which or what's the organization's name a specific guy is affiliated to.
        # When/During/when is start time ...
        # Etc.
        # If there is a statement from beginning of time to the end of time,
        # this will mean it is always true for the whole timeline.
        response = client.chat.completions.create(
            model=model_name,
            messages=[
                {
                    "role": "system",
                    "content": """You are an expert on paraphrasing questions.
                                  You job is paraphrasing this into a natural language question.
                                  The missing part can be someone, some organisation or some time.
                                  Use diverse ways to represent the temporal aspect of the question.
                                  Only return the paraphrased question, nothing else.
                                  """,
                },
                {
                    "role": "user",
                    "content": prompt_text,
                },
            ],
            max_tokens=100,
            temperature=0.8,
            stop=["\n"],
        )
        paraphrased_question = response.choices[0].message.content
        return paraphrased_question
    except Exception as e:
        print(f"An error occurred: {e}")
        return ""