openclaw-voc-skill/review-analysis/review-sentiment-analysis/api-config.json

{
  "name": "review-sentiment-analysis",
  "displayName": "评论情感分析",
  "description": "对评论语料库进行情感分析，分类正面/负面/中性，提取情感关键短语并按ASIN/品牌聚合",
  "category": "review-analysis",
  "version": "1.3.0",
  "type": "analysis",
  "parameters": {
    "reviewCorpus": {
      "type": "object",
      "required": true,
      "description": "review-batch-collection 输出的评论语料"
    },
    "language": {
      "type": "string",
      "required": false,
      "default": "en",
      "description": "评论语言"
    },
    "sentimentThreshold": {
      "type": "object",
      "required": false,
      "default": { "positiveMin": 0.6, "negativeMax": 0.4 },
      "properties": {
        "positiveMin": { "type": "number" },
        "negativeMax": { "type": "number" }
      },
      "description": "情感分类阈值"
    }
  },
  "pipeline": [
    {
      "step": 1,
      "name": "星级情感分类",
      "type": "compute",
      "logic": "sentimentClassify(reviewCorpus.reviews, { language, sentimentThreshold })",
      "algorithm": {
        "primaryRule": "基于星级快速分类: star>=4→positive, star<=2→negative, star==3→neutral",
        "secondaryRule": "可选NLP情感分析微调，处理4★但内容消极/2★但内容积极的边缘情况",
        "implementation": "遍历reviews数组，根据star字段分类，边缘情况可选用NLP微调"
      },
      "output": "classifiedReviews"
    },
    {
      "step": 2,
      "name": "按ASIN聚合情感指标",
      "type": "compute",
      "logic": "aggregateByAsin(classifiedReviews)",
      "algorithm": {
        "perAsin": "{ positivePct, neutralPct, negativePct, total }",
        "percentCalc": "Math.round((count/total)*1000)/10 → 保留一位小数",
        "implementation": "groupBy(asin) → 每组统计positive/neutral/negative数量和百分比"
      },
      "output": "asinAggregation"
    },
    {
      "step": 3,
      "name": "提取全局情感热词",
      "type": "compute",
      "logic": "extractGlobalHotPhrases(classifiedReviews)",
      "algorithm": {
        "tokenize": "分词 + 停用词过滤(the/a/an/is/it/i/my/to/and/of/for/in...)",
        "freqCount": "统计词频，按频率降序排列",
        "phraseExtract": "提取高频短语(2-3词组合)，如'scent throw', 'oil refill'",
        "splitBysentiment": "分别统计正面/负面评论中的高频词",
        "implementation": "split+toLowerCase+停用词过滤 → Map<word,count> → 按count降序取topN"
      },
      "output": "globalHotPhrases"
    },
    {
      "step": 4,
      "name": "情感短语识别+情感词云",
      "type": "compute",
      "logic": "extractSentimentKeywordCloud(classifiedReviews)",
      "algorithm": {
        "sentimentPhraseRules": {
          "description": "Regex模式匹配多词情感短语，短语优先于单词",
          "negativePhrases": [
            "too small", "too big", "too short", "too long", "too tight", "too thin",
            "poor quality", "bad quality", "cheaply made",
            "fell apart", "doesn't fit", "not worth", "waste of money",
            "not as described/pictured/shown/advertised",
            "wrong size", "wrong color", "bad smell", "chemical smell",
            "sent back", "returned it", "ripped/torn", "stained"
          ],
          "positivePhrases": [
            "well made", "high quality", "great quality", "good quality",
            "perfect fit", "fits perfectly/great/well",
            "highly recommend", "fast shipping",
            "great value/price/deal", "worth the money/penny",
            "loved it", "very comfortable/soft/warm",
            "true to size", "exceeded expectations",
            "beautiful/gorgeous", "super soft", "pleasantly surprised"
          ]
        },
        "sentimentWordDictionary": {
          "positive": ["comfortable", "durable", "sturdy", "soft", "affordable", "flattering", "recommend", "amazing", "awesome", "excellent", "fantastic", "stylish", "elegant", "versatile", "lightweight", "breathable"],
          "negative": ["disappointed", "disappointing", "flimsy", "cheap", "broke", "broken", "defective", "damaged", "uncomfortable", "itchy", "scratchy", "rough", "terrible", "horrible", "awful", "worst", "misleading", "refund", "shrunk", "faded", "wrinkled", "unraveled", "fraying", "pilling", "transparent", "see-through", "smelly", "stiff"]
        },
        "mergeAndOutput": {
          "phraseMinCount": 2,
          "wordMinCount": 3,
          "perSentiment": "正面和负面各取top10",
          "weight": "count/maxCount * 100"
        }
      },
      "output": "sentimentKeywordCloud"
    }
  ],
  "response": {
    "type": "object",
    "properties": {
      "reviewSentiments": {
        "type": "array",
        "description": "带情感标签的评论列表",
        "items": {
          "type": "object",
          "properties": {
            "asin": { "type": "string" },
            "star": { "type": "integer" },
            "sentiment": { "type": "string", "enum": ["positive", "negative", "neutral"] },
            "title": { "type": "string" },
            "content": { "type": "string" }
          }
        }
      },
      "asinAggregation": {
        "type": "object",
        "description": "按ASIN聚合的情感指标",
        "additionalProperties": {
          "type": "object",
          "properties": {
            "total": { "type": "integer" },
            "positivePct": { "type": "number" },
            "neutralPct": { "type": "number" },
            "negativePct": { "type": "number" }
          }
        }
      },
      "globalHotPhrases": {
        "type": "object",
        "description": "全局情感热词",
        "properties": {
          "positive": { "type": "array", "items": { "type": "object", "properties": { "phrase": { "type": "string" }, "frequency": { "type": "integer" } } } },
          "negative": { "type": "array", "items": { "type": "object", "properties": { "phrase": { "type": "string" }, "frequency": { "type": "integer" } } } }
        }
      },
      "sentimentKeywordCloud": {
        "type": "array",
        "description": "带情感标签的关键词云(短语+单词)",
        "items": {
          "type": "object",
          "properties": {
            "text": { "type": "string" },
            "weight": { "type": "integer", "description": "0-100权重" },
            "sentiment": { "type": "string", "enum": ["positive", "negative"] }
          }
        }
      }
    }
  },
  "algorithmNotes": {
    "starRatings": "Sorftime返回的星级字段(fiveStartRatings等)是百分比值，不是计数，加权评分=sum(star*pct)/sum(pct)",
    "stopWords": "English停用词表: the/a/an/is/it/i/my/to/and/of/for/in/this/that/was/with/but/have/not/are"
  },
  "timeout": 300000,
  "retry": {
    "maxAttempts": 2,
    "delay": 1000,
    "backoffMultiplier": 2
  }
}