Sentiment Shift Detector

Identify stocks where blogger sentiment has changed significantly.

Triggers

• •"情绪变化最大的股票"
• •"谁转向了"
• •"sentiment shift"
• •"who changed their mind"
• •"态度转变"
• •/sentiment-shift

Instructions

When the user wants to find sentiment shifts, follow these steps:

1. •

   Get Multi-Day Summaries Call get_daily_summary for recent dates (today, yesterday, a few days ago) to compare sentiment over time.

2. •

   Identify Changed Tickers Compare the summaries to find:

   • •Stocks that moved from bullish to bearish
   • •Stocks that moved from bearish to bullish
   • •Stocks with increased/decreased mentions
3. •

   Get Detailed Sentiment For tickers with notable changes, call get_ticker_sentiment to understand who changed their view.

4. •

   Search for Explanations Call search_viewpoints for changed tickers to find the reasoning behind sentiment shifts.

5. •

   Present Results Format the output as:

   code

   ## 情绪转变追踪 🔄
   
   ### 转向看涨 📈
   
   #### TICKER1
   - **变化**: 看跌 → 看涨
   - **时间**: X天前开始转变
   - **关键转变博主**: 博主A, 博主B
   - **转变原因**: [摘要为什么改变看法]
   - **代表观点**: "[具体观点]" — 博主A
   
   ### 转向看跌 📉
   
   #### TICKER2
   - **变化**: 看涨 → 看跌
   - **时间**: X天前开始转变
   - **关键转变博主**: 博主C
   - **转变原因**: [摘要为什么改变看法]
   - **代表观点**: "[具体观点]" — 博主C
   
   ### 热度变化 🌡️
   
   | 股票 | 之前提及 | 现在提及 | 变化 |
   |------|----------|----------|------|
   | XXX | 5 | 25 | ⬆️ +400% |
   | YYY | 20 | 3 | ⬇️ -85% |
   
   ### 分析
   [总结市场情绪变化的整体趋势]
   

Tool Sequence

1. •get_daily_summary(date=today) + get_daily_summary(date=yesterday) + get_daily_summary(date=3_days_ago) → Compare over time
2. •Identify tickers with changed sentiment
3. •get_ticker_sentiment(changed_ticker) → For each changed ticker
4. •search_viewpoints(changed_ticker) → Find reasoning
5. •Compile shift analysis

Notes

• •Sentiment shifts can be leading indicators
• •A blogger changing their view is often more significant than new bloggers joining
• •Track both direction changes and intensity changes
• •Include context for why shifts happened (news, earnings, etc.)