100个爆款关键词

import re
import jieba
import numpy as np
from collections import Counter
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
import synonyms  # 中文近义词库

class KeywordGenerator:
    def __init__(self):
        # 初始化停用词和种子词库
        self.stopwords = self._load_stopwords()
        self.seed_words = self._load_seed_words()
        jieba.initialize()

    def _load_stopwords(self):
        """加载停用词表"""
        with open('stopwords.txt', 'r', encoding='utf-8') as f:
            return set([line.strip() for line in f])

    def _load_seed_words(self):
        """加载行业种子词"""
        return {
            '电商': ['爆款', '限时', '折扣', '热卖'],
            '美妆': ['口红', '护肤', '彩妆', '美颜'],
            # 其他行业...
        }

    def generate_keywords(self, base_words, industry=None, num=100):
        """
        生成优化后的关键词
        :param base_words: 基础关键词列表
        :param industry: 所属行业
        :param num: 需要生成的数量
        :return: 排序后的关键词列表(带评分)
        """
        # 1. 关键词扩展
        expanded = self._expand_keywords(base_words, industry)

        # 2. 排重处理
        deduped = self._deduplicate_keywords(expanded)

        # 3. 质量评分
        scored_keywords = []
        for kw in deduped:
            score = self._calculate_quality_score(kw, industry)
            scored_keywords.append((kw, score))

        # 4. 排序并返回TopN
        scored_keywords.sort(key=lambda x: x[1], reverse=True)
        return [item[0] for item in scored_keywords[:num]]

    def _expand_keywords(self, base_words, industry):
        """关键词扩展(加入行业词、近义词等)"""
        expanded = set(base_words)

        # 加入行业种子词
        if industry in self.seed_words:
            expanded.update(self.seed_words[industry])

        # 近义词扩展
        for word in base_words:
            try:
                syns = synonyms.nearby(word)[0][:3]  # 每个词取3个近义词
                expanded.update(syns)
            except:
                continue

        return list(expanded)

    def _deduplicate_keywords(self, keywords):
        """基于语义相似度去重"""
        # 1. 文本向量化
        vectorizer = TfidfVectorizer(tokenizer=jieba.cut)
        tfidf_matrix = vectorizer.fit_transform(keywords)

        # 2. 计算相似度矩阵
        similarity_matrix = cosine_similarity(tfidf_matrix)

        # 3. 标记需要删除的重复项
        to_remove = set()
        for i in range(len(keywords)):
            for j in range(i+1, len(keywords)):
                if similarity_matrix[i][j] > 0.8:  # 相似度阈值
                    # 保留更短的关键词(通常更通用)
                    if len(keywords[i]) >= len(keywords[j]):
                        to_remove.add(i)
                    else:
                        to_remove.add(j)

        # 4. 返回去重后的列表
        return [kw for idx, kw in enumerate(keywords) if idx not in to_remove]

    def _calculate_quality_score(self, keyword, industry):
        """关键词质量评分算法"""
        score = 0

        # 1. 基础评分(长度、特殊符号等)
        length = len(keyword)
        if 4 <= length <= 8:  # 理想长度
            score += 2
        elif 8 < length <= 12:
            score += 1

        if re.search(r'[!！？?]', keyword):  # 带情感符号
            score += 0.5

        # 2. 行业相关性
        if industry:
            industry_words = self.seed_words.get(industry, [])
            for word in industry_words:
                if word in keyword:
                    score += 1.5
                    break

        # 3. 流行度指标(词频统计)
        words = [w for w in jieba.cut(keyword) if w not in self.stopwords]
        freq_score = sum([self._get_word_frequency(w) for w in words]) / len(words) if words else 0
        score += freq_score * 2

        # 4. 多样性评分(词性组合)
        pos_tags = [pair[1] for pair in jieba.posseg.cut(keyword)]
        unique_pos = len(set(pos_tags))
        score += unique_pos * 0.8

        return round(score, 2)

    def _get_word_frequency(self, word):
        """获取词频(模拟实现)"""
        # 实际应用中可以从搜索日志或语料库统计
        freq_dict = {
            '爆款': 0.9, '折扣': 0.8, '口红': 0.7, 
            '限时': 0.6, '热卖': 0.5, '护肤': 0.4
        }
        return freq_dict.get(word, 0.3)

    def evaluate_keywords(self, keywords):
        """批量评估关键词质量"""
        return [(kw, self._calculate_quality_score(kw)) for kw in keywords]


# 使用示例
if __name__ == "__main__":
    generator = KeywordGenerator()

    # 基础关键词
    base_keywords = ["爆款口红", "限时折扣", "美妆护肤", "热卖商品"]

    # 生成优化后的关键词
    optimized_keywords = generator.generate_keywords(
        base_keywords, 
        industry="美妆",
        num=50
    )

    print("优化后的关键词(带评分):")
    for kw, score in generator.evaluate_keywords(optimized_keywords[:20]):
        print(f"{kw} (评分: {score})")