fmode-voc-projects/scripts/tools/hongchengdaojia/hongcheng-analyze.js

// ==============================================================================
// 洪城到家 · VOC 数据分析模块
// ==============================================================================

const fs = require('fs');
const path = require('path');

const ROOT = path.resolve(__dirname, '..', '..');
const RAW_DIR = path.join(ROOT, 'docs', '洪城到家', 'raw');
const MERGED_PATH = path.join(RAW_DIR, '_merged.json');

const HYPOTHESES = {
    H1: { title: '医院地推是最高效获客场景', desc: '直接触达即将生产的精准用户，转化率远高于泛流量' },
    H2: { title: '价格透明是转化的关键因素', desc: '用户对月嫂价格不透明是核心痛点，价格公开透明化能显著提升转化率' },
    H3: { title: '短剧内容能有效提升品牌认知', desc: '借南昌万亿短剧市场政策红利，通过短剧植入能低成本实现品牌曝光' },
    H4: { title: '专业度信任需要"看得见"的证明', desc: '用户判断月嫂专业度的核心信号是资质证书、客户评价、实体规模' },
    H5: { title: '竞品对比中社区店是主要威胁', desc: '社区月嫂门店对洪诚到家的主要威胁是"价格低"，而非"质量好"' },
    H6: { title: '老带新是低成本高转化渠道', desc: '老客户转介绍是月嫂业务最高效、最省钱的获客方式' },
    H7: { title: '美团/小红书是搜索承接的主力平台', desc: '用户在网络上主动搜索月嫂的主要平台是美团和小红书' },
    H8: { title: '"不满意能换"是重要的服务保障', desc: '用户担心"月嫂临时变更或服务不达标"，服务保障承诺能显著降低决策门槛' },
};

const CHAPTER_HYPOTHESIS_MAP = {
    challenge: ['H1', 'H2', 'H4'],
    competitor: ['H1', 'H5', 'H7'],
    pain: ['H2', 'H4', 'H8'],
    kano: ['H2', 'H4', 'H6'],
    scene: ['H1', 'H3'],
    threeaxis: ['H5', 'H7'],
    opportunity: ['H1', 'H3', 'H7'],
    blueprint: ['H2', 'H4', 'H7', 'H8'],
};

const PLATFORM_LABELS = {
    xhs: { name: '小红书', color: '#FF2442', short: '红' },
    douyin: { name: '抖音', color: '#1A1A1A', short: '抖' },
    meituan: { name: '美团', color: '#FFD100', short: '美' },
    unknown: { name: '其他', color: '#888', short: '—' },
};

const KEYWORD_TO_PRODUCT = {
    '洪诚到家': '洪诚到家',
    '洪城到家': '洪诚到家',
    '南昌月嫂洪诚': '洪诚到家',
    '洪诚到家怎么样': '洪诚到家（口碑）',
    '天鹅到家月嫂': '天鹅到家',
    '好孕妈妈月嫂': '好孕妈妈',
    '多喜娃月嫂': '多喜娃',
    '妈咪无忧月嫂': '妈咪无忧',
    '新手爸妈找月嫂': '场景话题',
    '坐月子注意事项': '场景话题',
    '月嫂多少钱一个月': '价格话题',
    '婆婆不让请月嫂': '场景话题',
    '月嫂怎么找靠谱': '决策话题',
    '月嫂不满意可以换吗': '服务保障话题',
    '社区月嫂': '社区月嫂（竞品）',
    '南昌月嫂对比': '竞品对比',
    '朋友推荐月嫂': '口碑话题',
    '南昌月嫂价格': '价格话题',
    '月嫂服务': '品类话题',
    '职场妈妈月子': '场景话题',
    '新手爸妈': '场景话题',
    '南昌短剧': '短剧话题',
    '婆媳关系坐月子': '场景话题',
    '南昌月嫂': '品类词',
    '南昌月嫂哪家好': '品类词',
    '江西月嫂': '品类词',
};

const NEG_KEYS = ['差', '坑', '骗', '贵', '不专业', '不满意', '后悔', '吐槽', '失望', '糟糕', '没用', '没效果', '不靠谱', '坑人'];
const POS_KEYS = ['好', '推荐', '满意', '专业', '靠谱', '放心', '值得', '不错', '棒', '优秀', '喜欢', '感谢', '回购'];
const CONF_KEYS = ['但是', '可是', '纠结', '担心', '犹豫', '想又怕', '不知道.*好'];

function inferSentiment(item) {
    if (item.sentiment && ['positive', 'negative', 'neutral', 'conflicted'].includes(item.sentiment)) return item.sentiment;
    const c = String(item.content || '').toLowerCase();
    for (const k of NEG_KEYS) if (c.includes(k.toLowerCase())) return 'negative';
    for (const k of POS_KEYS) if (c.includes(k.toLowerCase())) return 'positive';
    for (const k of CONF_KEYS) if (new RegExp(k).test(c)) return 'conflicted';
    return 'neutral';
}

const SKELETON = {
    meta: {
        collectedAt: '待采集',
        platforms: {},
        hypotheses: {},
        products: {},
        stage: 'skeleton',
        sourceNote: '暂无采集数据',
    },
    items: [],
    raw: null,
};

function loadMerged() {
    if (fs.existsSync(MERGED_PATH)) {
        try {
            const raw = JSON.parse(fs.readFileSync(MERGED_PATH, 'utf8'));
            const items = (raw.items || []).map((it) => ({
                ...it,
                product: KEYWORD_TO_PRODUCT[it.keyword] || it.product || it.keyword,
                sentiment: inferSentiment(it),
            }));
            return {
                meta: {
                    sourceTier: raw.meta?.sourceTier || 'real-collected',
                    stage: raw.meta?.stage || 'batch-real',
                    collectedAt: raw.meta?.collectedAt || new Date().toISOString().slice(0, 10),
                    product: '洪城到家',
                    stats: raw.meta || {},
                    sourceNote: 'docs/洪城到家/raw/_merged.json · 真实多平台采集',
                },
                items,
                raw,
            };
        } catch (err) {
            console.warn(`⚠ _merged.json 解析失败：${err.message}`);
        }
    }
    return SKELETON;
}

function getMeta(data) {
    const items = (data && data.items) || [];
    const platforms = {};
    const products = {};
    const hypotheses = {};
    const sources = {};
    const sentiments = {};
    const keywords = new Set();
    const tags = new Set();

    for (const it of items) {
        const pf = it.platform || 'unknown';
        platforms[pf] = (platforms[pf] || 0) + 1;
        const prod = it.product || 'unknown';
        products[prod] = (products[prod] || 0) + 1;
        const hs = Array.isArray(it.hypothesis) ? it.hypothesis : (it.hypothesis ? [it.hypothesis] : []);
        for (const h of hs) hypotheses[h] = (hypotheses[h] || 0) + 1;
        const src = it.source || 'unknown';
        sources[src] = (sources[src] || 0) + 1;
        sentiments[it.sentiment || 'unknown'] = (sentiments[it.sentiment || 'unknown'] || 0) + 1;
        if (it.keyword) keywords.add(it.keyword);
        if (Array.isArray(it.tags)) it.tags.forEach((t) => tags.add(t));
    }

    return {
        comments: items.length,
        keywords: keywords.size,
        tagsTotal: tags.size,
        platforms,
        products,
        productsCount: Object.keys(products).length,
        hypotheses,
        sources,
        sentiments,
        stage: data?.meta?.stage || 'unknown',
        sourceTier: data?.meta?.sourceTier || 'unknown',
        collectedAt: data?.meta?.collectedAt || '待采集',
        sourceNote: data?.meta?.sourceNote || '',
    };
}

function filterByHypothesis(items, h) {
    return items.filter((it) => {
        const hs = Array.isArray(it.hypothesis) ? it.hypothesis : (it.hypothesis ? [it.hypothesis] : []);
        return hs.includes(h);
    });
}
function filterByProduct(items, product) {
    return items.filter((it) => (it.product || '').includes(product) || (it.keyword || '').includes(product));
}
function filterByKeyword(items, kw) {
    return items.filter((it) => it.keyword === kw);
}
function filterByPlatform(items, platform) {
    return items.filter((it) => it.platform === platform);
}
function filterBySentiment(items, sentiment) {
    return items.filter((it) => it.sentiment === sentiment);
}
function filterByTag(items, tag) {
    return items.filter((it) => Array.isArray(it.tags) && it.tags.some((t) => t.includes(tag)));
}
function filterByContent(items, re) {
    const rx = re instanceof RegExp ? re : new RegExp(String(re), 'i');
    return items.filter((it) => rx.test(String(it.content || '')));
}
function filterByMinLikes(items, min = 1) {
    return items.filter((it) => (it.likes || 0) >= min);
}

function topByLikes(items, n = 10) {
    return items.slice().sort((a, b) => (b.likes || 0) - (a.likes || 0)).slice(0, n);
}
function sample(items, n = 6, seed = 1) {
    const arr = items.slice();
    const result = [];
    let s = seed;
    while (result.length < n && arr.length) {
        s = (s * 9301 + 49297) % 233280;
        const idx = Math.floor((s / 233280) * arr.length);
        result.push(arr.splice(idx, 1)[0]);
    }
    return result;
}
function groupByTag(items) {
    const map = new Map();
    for (const it of items) {
        if (!Array.isArray(it.tags)) continue;
        for (const t of it.tags) {
            if (!map.has(t)) map.set(t, { tag: t, count: 0, items: [] });
            const g = map.get(t);
            g.count++;
            g.items.push(it);
        }
    }
    return Array.from(map.values()).sort((a, b) => b.count - a.count);
}

function isSubstantive(content, minChars) {
    const s = String(content || '').trim();
    if (s.length < minChars) return false;
    const stripped = s.replace(/\[[^\]]+\]/g, '').replace(/[\s\p{P}\p{Emoji_Presentation}\p{Extended_Pictographic}]/gu, '');
    return stripped.length >= Math.max(4, Math.floor(minChars / 2));
}

function getEvidence(items, opts = {}) {
    const {
        hypothesis, product, keyword, platform, sentiment, tag,
        minLikes = 0, minChars = 10, contentMatch, requireContentHit = false,
        dedupByContent = true, dedupByNickname = false,
        top = 6, seed = 7, sortBy = 'likes',
    } = opts;
    let filtered = items.slice();
    if (hypothesis) filtered = filterByHypothesis(filtered, hypothesis);
    if (product) filtered = filterByProduct(filtered, product);
    if (keyword) filtered = filterByKeyword(filtered, keyword);
    if (platform) filtered = filterByPlatform(filtered, platform);
    if (sentiment) filtered = filterBySentiment(filtered, sentiment);
    if (tag) filtered = filterByTag(filtered, tag);
    if (minLikes) filtered = filterByMinLikes(filtered, minLikes);
    filtered = filtered.filter((it) => isSubstantive(it.content, minChars));
    if (contentMatch) filtered = filterByContent(filtered, contentMatch);

    if (sortBy === 'likes') {
        filtered = filtered.sort((a, b) => (b.likes || 0) - (a.likes || 0));
    }

    if (dedupByContent) {
        const seen = new Set();
        filtered = filtered.filter((it) => {
            const key = dedupByNickname
                ? `${(it.content || '').slice(0, 40)}|${it.nickname || ''}`
                : (it.content || '').slice(0, 40);
            if (seen.has(key)) return false;
            seen.add(key);
            return true;
        });
    }

    const pool = filtered.slice(0, Math.max(top * 2, top + 3));
    return sample(pool, Math.min(top, pool.length), seed);
}

function isSeed(item) { return (item?.source || '').includes('pattern') || (item?.source || '').includes('seed'); }
function isReal(item) { return (item?.source || '') === 'real-collected'; }

function getGlobalSourceLabel(meta) {
    if (!meta) return '未加载';
    const tier = meta.sourceTier;
    if (tier === 'real-collected') return '真实采集';
    if (tier === 'pattern-curated') return '公开模式归纳 · 种子样本';
    return '骨架占位';
}

module.exports = {
    HYPOTHESES,
    CHAPTER_HYPOTHESIS_MAP,
    PLATFORM_LABELS,
    loadMerged,
    getMeta,
    filterByHypothesis, filterByProduct, filterByKeyword, filterByPlatform,
    filterBySentiment, filterByTag, filterByContent, filterByMinLikes,
    topByLikes, sample, groupByTag,
    getEvidence,
    isSeed, isReal,
    getGlobalSourceLabel,
};