Source code for data_juicer.ops.mapper.extract_nickname_mapper

import re
from typing import Dict, Optional

import numpy as np
from loguru import logger
from pydantic import PositiveInt

from data_juicer.ops.base_op import OPERATORS, TAGGING_OPS, Mapper
from data_juicer.utils.constant import Fields, MetaKeys
from data_juicer.utils.model_utils import get_model, prepare_model

OP_NAME = "extract_nickname_mapper"


# TODO: LLM-based inference.


@TAGGING_OPS.register_module(OP_NAME)
@OPERATORS.register_module(OP_NAME)
class ExtractNicknameMapper(Mapper):
    """
    Extract nickname relationship in the text.
    """

    DEFAULT_SYSTEM_PROMPT = (
        "给定你一段文本，你的任务是将人物之间的称呼方式（昵称）提取出来。\n"
        "要求：\n"
        "- 需要给出说话人对被称呼人的称呼，不要搞反了。\n"
        "- 相同的说话人和被称呼人最多给出一个最常用的称呼。\n"
        "- 请不要输出互相没有昵称的称呼方式。\n"
        "- 输出格式如下：\n"
        "```\n"
        "### 称呼方式1\n"
        "- **说话人**：...\n"
        "- **被称呼人**：...\n"
        "- **...对...的昵称**：...\n"
        "### 称呼方式2\n"
        "- **说话人**：...\n"
        "- **被称呼人**：...\n"
        "- **...对...的昵称**：...\n"
        "### 称呼方式3\n"
        "- **说话人**：...\n"
        "- **被称呼人**：...\n"
        "- **...对...的昵称**：...\n"
        "...\n"
        "```\n"
    )
    DEFAULT_INPUT_TEMPLATE = "# 文本\n```\n{text}\n```\n"
    DEFAULT_OUTPUT_PATTERN = r"""
        \#\#\#\s*称呼方式(\d+)\s*
        -\s*\*\*说话人\*\*\s*：\s*(.*?)\s*
        -\s*\*\*被称呼人\*\*\s*：\s*(.*?)\s*
        -\s*\*\*(.*?)对(.*?)的昵称\*\*\s*：\s*(.*?)(?=\#\#\#|\Z) # for double check
    """



    def __init__(
        self,
        api_model: str = "gpt-4o",
        *,
        nickname_key: str = MetaKeys.nickname,
        api_endpoint: Optional[str] = None,
        response_path: Optional[str] = None,
        system_prompt: Optional[str] = None,
        input_template: Optional[str] = None,
        output_pattern: Optional[str] = None,
        try_num: PositiveInt = 3,
        drop_text: bool = False,
        model_params: Dict = {},
        sampling_params: Dict = {},
        **kwargs,
    ):
        """
        Initialization method.
        :param api_model: API model name.
        :param nickname_key: The key name to store the nickname
            relationship in the meta field. It's "nickname" in default.
        :param api_endpoint: URL endpoint for the API.
        :param response_path: Path to extract content from the API response.
            Defaults to 'choices.0.message.content'.
        :param system_prompt: System prompt for the task.
        :param input_template: Template for building the model input.
        :param output_pattern: Regular expression for parsing model output.
        :param try_num: The number of retry attempts when there is an API
            call error or output parsing error.
        :param drop_text: If drop the text in the output.
        :param model_params: Parameters for initializing the API model.
        :param sampling_params: Extra parameters passed to the API call.
            e.g {'temperature': 0.9, 'top_p': 0.95}
        :param kwargs: Extra keyword arguments.
        """
        super().__init__(**kwargs)

        self.nickname_key = nickname_key

        self.system_prompt = system_prompt or self.DEFAULT_SYSTEM_PROMPT
        self.input_template = input_template or self.DEFAULT_INPUT_TEMPLATE
        self.output_pattern = output_pattern or self.DEFAULT_OUTPUT_PATTERN

        self.sampling_params = sampling_params
        self.model_key = prepare_model(
            model_type="api", model=api_model, endpoint=api_endpoint, response_path=response_path, **model_params
        )

        self.try_num = try_num
        self.drop_text = drop_text




    def parse_output(self, raw_output):
        pattern = re.compile(self.output_pattern, re.VERBOSE | re.DOTALL)
        matches = pattern.findall(raw_output)

        nickname_relations = []

        for match in matches:
            _, role1, role2, role1_tmp, role2_tmp, nickname = match
            # for double check
            if role1.strip() != role1_tmp.strip() or role2.strip() != role2_tmp.strip():
                continue
            role1 = role1.strip()
            role2 = role2.strip()
            nickname = nickname.strip()
            # is name but not nickname
            if role2 == nickname:
                continue
            if role1 and role2 and nickname:
                nickname_relations.append((role1, role2, nickname))
        nickname_relations = list(set(nickname_relations))

        nickname_relations = [
            {
                MetaKeys.source_entity: nr[0],
                MetaKeys.target_entity: nr[1],
                MetaKeys.relation_description: nr[2],
                MetaKeys.relation_keywords: ["nickname"],
                MetaKeys.relation_strength: None,
            }
            for nr in nickname_relations
        ]

        return nickname_relations




    def process_single(self, sample, rank=None):
        # check if it's generated already
        if self.nickname_key in sample[Fields.meta]:
            return sample

        client = get_model(self.model_key, rank=rank)

        input_prompt = self.input_template.format(text=sample[self.text_key])
        messages = [{"role": "system", "content": self.system_prompt}, {"role": "user", "content": input_prompt}]
        nickname_relations = [
            {
                MetaKeys.source_entity: "",
                MetaKeys.target_entity: "",
                MetaKeys.relation_description: "",
                MetaKeys.relation_keywords: np.array([], dtype=str),
                MetaKeys.relation_strength: None,
            }
        ]
        for _ in range(self.try_num):
            try:
                output = client(messages, **self.sampling_params)
                results = self.parse_output(output)
                if len(results) > 0:
                    nickname_relations = results
                    break
            except Exception as e:
                logger.warning(f"Exception: {e}")

        sample[Fields.meta][self.nickname_key] = nickname_relations
        if self.drop_text:
            sample.pop(self.text_key)

        return sample