SuperAPI AI Directory

unified-comment-analysis

Marketing & growth

View repository

Updated 10-15 19:55

About this skill

Multi-platform public opinion monitoring and operational insight dashboard

Overview

Activate this skill when the user provides a share link from Douyin, Xiaohongshu, Bilibili, or Kuaishou and needs to analyze comment sentiment, user profiles, conversion potential, or operational recommendations.

This skill calls the backend API through a 3-step asynchronous workflow (parse-link -> create-task -> get-task) to obtain comment sentiment analysis, user profiles, conversion potential assessment, and operational recommendations.

Invocation Mode

This skill uses comment-analysis-task asynchronous workflow, which includes the following three steps:

  1. parse-link (POST /api/v1/comment-analysis/parse-link): Extract the platform type, content ID, and title from the share link
  2. create-task (POST /api/v1/comment-analysis/tasks): Submit the public opinion analysis task and obtain the task ID
  3. get-task (GET /api/v1/comment-analysis/tasks/:id): Poll the task status until completion (up to 60 retries, 2-second interval)

Parameters

See references/form-schema.md for the full parameter schema.

Parameter Type Required Description
link string Yes Share link, in URI format (Douyin/Xiaohongshu/Bilibili/Kuaishou)
platform enum No Platform type, optional values:douyin(Douyin),xhs(Xiaohongshu),bilibili(Bilibili),kuaishou(Kuaishou). If not provided, it will be automatically identified from the link

Execution

At runtime, it is executed through scripts/run.py , and internally completes the 3-step asynchronous workflow automatically:

python3 scripts/run.py --params '{"link":"https://www.douyin.com/video/7381234567890123456"}'

Optionally specify the platform:

python3 scripts/run.py --params '{"link":"https://example.com/xxx","platform":"douyin"}'

3-step asynchronous workflow details

Step 1: parse-link

POST /api/v1/comment-analysis/parse-link
Content-Type: application/json

{
  "input": "<分享链接>",
  "platform": "<可选平台>"  // 若用户未指定则由系统自动识别
}

Response example:

{
  "success": true,
  "data": {
    "platform": "douyin",
    "contentId": "7381234567890123456",
    "contentTitle": "视频标题",
    "sourceUrl": "https://www.douyin.com/video/7381234567890123456"
  }
}

Step 2: create-task

POST /api/v1/comment-analysis/tasks
Content-Type: application/json

{
  "platform": "douyin",
  "contentId": "7381234567890123456",
  "contentTitle": "视频标题",
  "options": {
    "sourceUrl": "https://www.douyin.com/video/7381234567890123456"
  }
}

Response example:

{
  "success": true,
  "data": {
    "task": {
      "id": "task_abc123"
    }
  }
}

Step 3: get-task (polling)

GET /api/v1/comment-analysis/tasks/:id

Polling logic (built into scripts/run.py):

  • Poll every 2 seconds
  • Up to 60 retries
  • Status is completed or failed to stop polling

Required environment variables:

Environment variable Description
AISKILLS_BASE_URL API base address (default:https://ai-skills.ai)
AISKILLS_API_KEY API key (required)
AISKILLS_TENANT_ID Tenant ID (default:default)

Output Format

The returned results are organized according to the conclusion -> evidence -> recommendation structure. For details, see references/output-format.md。

Core returned fields:

Field Description
sentiment Sentiment analysis: positive / negative / neutral
sentimentRatio Sentiment proportions (proportion of positive/negative/neutral comments)
userProfile User profile: age group, interest tags, geographic distribution
conversionPotential Conversion potential score (0-100)
engagementMetrics Interaction metrics: likes, comments, shares, favorites
suggestions List of operational recommendations

Examples

Example 1: Douyin video comment analysis

Input:

Help me analyze the comments on this Douyin video:https://www.douyin.com/video/7381234567890123456

Output:

结论:该视频整体舆情偏正面(正面评论占比 68%),用户互动积极,但负面评论集中在视频后半段。

证据:
- 情感分析:正面 68%,中性 22%,负面 10%
- 评论峰值出现在视频 0-30 秒,正面情绪占比高达 75%
- 负面评论多在 2:15 处(涉及争议性话题)
- 转化潜力评分:78/100(中等偏高)

用户画像:
- 主要受众:18-25 岁女性(占比 62%)
- 高热词:#春季穿搭 #平价好物 #学生党
- 地域分布:广东(18%)、浙江(14%)、江苏(11%)

建议:
1. 视频后半段加入正向引导,降低负面情绪扩散风险
2. 结合"学生党平价好物"标签制作后续内容,延续热度
3. 在评论区置顶"学生党专属优惠"引导评论互动

Example 2: Xiaohongshu note public opinion analysis

Input:

This is the link to my Xiaohongshu note. I want to see how everyone is reacting:https://www.xiaohongshu.com/explore/67890abcdef

Output:

结论:该笔记整体反响良好,正面情绪占主导(正面评论占比 74%),收藏率较高,用户认可度高。

证据:
- 情感分析:正面 74%,中性 19%,负面 7%
- 收藏率 15.3%(高于同类笔记均值 8.2%)
- 评论区高热词:#干货 #种草 #已入手
- 互动率:点赞 2.3w,收藏 8,600,评论 1,200

用户画像:
- 主要受众:25-35 岁都市女性(占比 71%)
- 兴趣标签:护肤、美妆、生活方式
- 地域分布:上海(22%)、北京(18%)、成都(12%)

建议:
1. 趁热度发布笔记 Part2,延续"种草清单"系列
2. 在评论区置顶"粉丝专属折扣码"提升转化
3. 私信高互动用户建立 KOC 关系链

Example 3: Bilibili video comment insights

Input:

Help me check the comment response to this Bilibili video:https://www.bilibili.com/video/BV1xx411c7mD

Output:

结论:该视频用户反馈整体正面(正面评论占比 71%),弹幕活跃度高,社区氛围良好。

证据:
- 情感分析:正面 71%,中性 21%,负面 8%
- 弹幕密度:平均 120 条/分钟(同类均值 85 条)
- 评论峰值:视频 5:00-8:00(干货密集区)
- 转化潜力评分:82/100(高)

用户画像:
- 主要受众:18-30 岁男性(占比 65%)
- 高热词:#硬核 #干货 #涨知识 #三连
- 互动特点:投币率 12%,三连率 18%

建议:
1. 内容系列化:拆分为"入门篇/进阶篇/实战篇"
2. 视频中植入相关课程/工具推荐,引导转化
3. 建立粉丝群维护高质量用户关系

Example 4: Kuaishou livestream short video public opinion

Input:

This is the link to my Kuaishou video. Help me analyze the comments:https://www.kuaishou.com/video/3x4c7e9f2g1h

Output:

结论:该视频评论以正面为主(正面占比 62%),但负面反馈需关注,互动率整体中等。

证据:
- 情感分析:正面 62%,中性 24%,负面 14%
- 负面评论集中在"价格偏高"和"物流速度"方面
- 转化潜力评分:65/100(有提升空间)

用户画像:
- 主要受众:25-45 岁下沉市场用户(占比 78%)
- 价格敏感度:高
- 地域分布:山东(19%)、河南(17%)、河北(15%)

建议:
1. 针对负面反馈(价格/物流)在评论区公开回应,展示服务态度
2. 推出限时优惠券或满减活动,提升下单转化率
3. 与粉丝建立更紧密互动(如直播连麦、粉丝见面会)