Spaces:
Running
Running
File size: 11,902 Bytes
5d2eba0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 | """
features/earnings_sentiment.py โ Earnings Call Sentiment Intelligence
Analyzes earnings call transcripts for sentiment, confidence, guidance tone.
"""
import streamlit as st
import json
import re
import logging
from datetime import datetime, timedelta
logger = logging.getLogger("EarningsSentiment")
# ---------------------------------------------------------------------------
# Transcript fetching
# ---------------------------------------------------------------------------
def _fetch_transcript(ticker: str, quarter: int = None, year: int = None) -> str:
"""Fetch earnings call transcript via Tavily search."""
from features.utils import run_tavily_search
now = datetime.now()
if not quarter:
quarter = (now.month - 1) // 3 + 1
# Last quarter
if quarter == 1:
quarter = 4
year = (year or now.year) - 1
else:
quarter -= 1
if not year:
year = now.year
query = f"{ticker} earnings call transcript Q{quarter} {year}"
try:
result = run_tavily_search(query, search_depth="advanced")
texts = []
for qr in result.get("data", []):
for r in qr.get("results", []):
texts.append(r.get("content", ""))
return "\n\n".join(texts[:5]) if texts else ""
except Exception as e:
logger.error(f"Transcript fetch failed: {e}")
return ""
# ---------------------------------------------------------------------------
# Sentiment analysis via Gemini
# ---------------------------------------------------------------------------
def _analyze_sentiment(ticker: str, transcript: str) -> dict:
"""Run Gemini to analyze sentiment of earnings call."""
from features.utils import call_gemini
prompt = f"""You are an expert sentiment analyst specializing in earnings.
Analyze the following text regarding the earnings call for {ticker}.
Note: The text may be the raw transcript OR market commentary/news about the call.
Analyze whatever is provided to determine the sentiment, guidance, and key themes as accurately as possible.
---
{transcript[:6000]}
---
Provide your analysis as a VALID JSON object with this exact structure:
{{
"management_sentiment": {{
"score": <float from -1.0 to 1.0>,
"label": "Positive" | "Neutral" | "Negative",
"confidence_level": <int from 0-100>,
"forward_guidance": "Optimistic" | "Cautious" | "Withdrawn",
"key_quotes": ["quote1", "quote2"]
}},
"qa_sentiment": {{
"score": <float from -1.0 to 1.0>,
"label": "Positive" | "Neutral" | "Negative",
"confidence_level": <int from 0-100>,
"analyst_concerns": ["concern1", "concern2"]
}},
"key_themes": ["theme1", "theme2", "theme3", "theme4", "theme5"],
"positive_words": ["word1", "word2", "word3", "word4", "word5", "word6", "word7", "word8"],
"negative_words": ["word1", "word2", "word3", "word4", "word5", "word6", "word7", "word8"],
"divergence_alerts": ["alert1 if any"],
"between_the_lines": "A 2-3 paragraph analysis of what management is really communicating between the lines."
}}
Be precise with scores. Detect hedging language, overconfidence, and tone shifts.
Return ONLY the JSON, no markdown formatting."""
raw = call_gemini(prompt, "You are a senior NLP analyst at a hedge fund specializing in earnings call analysis.")
# Force JSON format cleanup if AI included markdown blocks
raw = raw.replace("```json", "").replace("```", "").strip()
try:
# Match from first { to last }
json_match = re.search(r'\{.*\}', raw, re.DOTALL)
if json_match:
return json.loads(json_match.group(0))
except (json.JSONDecodeError, ValueError) as e:
logger.error(f"Sentiment parse error: {e}")
# Fallback structure
return {
"management_sentiment": {"score": 0, "label": "Neutral", "confidence_level": 50, "forward_guidance": "Cautious", "key_quotes": []},
"qa_sentiment": {"score": 0, "label": "Neutral", "confidence_level": 50, "analyst_concerns": []},
"key_themes": ["Unable to parse"],
"positive_words": [], "negative_words": [],
"divergence_alerts": [],
"between_the_lines": raw,
}
# ---------------------------------------------------------------------------
# Visualization helpers
# ---------------------------------------------------------------------------
def _render_gauge(score: float, label: str, title: str):
"""Render a Plotly gauge chart for sentiment score."""
import plotly.graph_objects as go
color = "#10b981" if score > 0.2 else "#ef4444" if score < -0.2 else "#f59e0b"
fig = go.Figure(go.Indicator(
mode="gauge+number+delta",
value=score,
title={"text": title, "font": {"size": 16, "color": "white"}},
number={"font": {"color": "white"}},
gauge={
"axis": {"range": [-1, 1], "tickcolor": "white"},
"bar": {"color": color},
"bgcolor": "#1e1e1e",
"bordercolor": "#333",
"steps": [
{"range": [-1, -0.3], "color": "rgba(239,68,68,0.2)"},
{"range": [-0.3, 0.3], "color": "rgba(245,158,11,0.2)"},
{"range": [0.3, 1], "color": "rgba(16,185,129,0.2)"},
],
},
))
fig.update_layout(
paper_bgcolor="rgba(0,0,0,0)", font_color="white",
height=250, margin=dict(l=20, r=20, t=50, b=20),
)
return fig
def _render_wordcloud(words: list, title: str, colormap: str = "Greens"):
"""Generate a word cloud image from a list of words."""
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
from wordcloud import WordCloud
if not words:
return None
text = " ".join(words)
wc = WordCloud(
width=400, height=200, background_color="black",
colormap=colormap, max_words=50, prefer_horizontal=0.7,
).generate(text)
fig, ax = plt.subplots(figsize=(6, 3))
ax.imshow(wc, interpolation="bilinear")
ax.axis("off")
ax.set_title(title, color="white", fontsize=12, pad=10)
fig.patch.set_facecolor("black")
plt.tight_layout()
return fig
# ---------------------------------------------------------------------------
# Streamlit page renderer
# ---------------------------------------------------------------------------
def render_earnings_sentiment():
st.markdown("## ๐๏ธ Earnings Call Sentiment Intelligence")
st.caption("Analyze earnings call transcripts for hidden sentiment signals, management confidence, "
"and forward guidance shifts that predict future price moves.")
col1, col2, col3 = st.columns([2, 1, 1])
with col1:
ticker = st.text_input("Ticker Symbol:", placeholder="e.g. AAPL", key="es_ticker").upper().strip()
with col2:
quarter = st.selectbox("Quarter:", [None, 1, 2, 3, 4], format_func=lambda x: f"Q{x}" if x else "Auto-detect", key="es_q")
with col3:
year = st.number_input("Year:", min_value=2020, max_value=2026, value=datetime.now().year, key="es_year")
analyze_btn = st.button("๐ Analyze Earnings Call", use_container_width=True, key="es_analyze")
if analyze_btn and ticker:
with st.status("๐๏ธ Analyzing earnings call...", expanded=True) as status:
status.write(f"๐ก Searching for {ticker} Q{quarter or 'latest'} {year} transcript...")
transcript = _fetch_transcript(ticker, quarter, year)
if not transcript:
status.update(label="โ ๏ธ No transcript found", state="error")
st.warning(f"Could not find earnings call transcript for {ticker}. "
"Try specifying a different quarter or year.")
return
status.write("๐ง Running deep sentiment analysis...")
analysis = _analyze_sentiment(ticker, transcript)
st.session_state["es_analysis"] = analysis
st.session_state["es_display_ticker"] = ticker
status.update(label="โ
Analysis Complete!", state="complete", expanded=False)
# Display results
analysis = st.session_state.get("es_analysis")
if not analysis:
return
ticker_display = st.session_state.get("es_display_ticker", "")
st.markdown(f"### ๐ Sentiment Analysis: **{ticker_display}**")
mgmt = analysis.get("management_sentiment", {})
qa = analysis.get("qa_sentiment", {})
# Side-by-side gauges
col1, col2 = st.columns(2)
with col1:
st.markdown("#### ๐ค Management Prepared Remarks")
fig = _render_gauge(mgmt.get("score", 0), mgmt.get("label", "N/A"), "Management Sentiment")
st.plotly_chart(fig, use_container_width=True)
st.markdown(f"**Confidence Level:** {mgmt.get('confidence_level', 'N/A')}/100")
st.markdown(f"**Forward Guidance:** {mgmt.get('forward_guidance', 'N/A')}")
if mgmt.get("key_quotes"):
st.markdown("**Key Quotes:**")
for q in mgmt["key_quotes"]:
st.markdown(f'> *"{q}"*')
with col2:
st.markdown("#### โ Q&A Session")
fig = _render_gauge(qa.get("score", 0), qa.get("label", "N/A"), "Q&A Sentiment")
st.plotly_chart(fig, use_container_width=True)
st.markdown(f"**Confidence Level:** {qa.get('confidence_level', 'N/A')}/100")
if qa.get("analyst_concerns"):
st.markdown("**Analyst Concerns:**")
for c in qa["analyst_concerns"]:
st.markdown(f"- โ ๏ธ {c}")
# Key Themes
st.markdown("---")
st.markdown("#### ๐ท๏ธ Key Themes Mentioned")
themes = analysis.get("key_themes", [])
if themes:
cols = st.columns(min(len(themes), 5))
for i, theme in enumerate(themes[:5]):
with cols[i % 5]:
st.markdown(f"""
<div style="background: #1e1e1e; border: 1px solid #333; border-radius: 8px;
padding: 12px; text-align: center; margin: 4px 0;">
<span style="font-size: 0.9rem; color: #a78bfa;">{theme}</span>
</div>
""", unsafe_allow_html=True)
# Word Clouds
col1, col2 = st.columns(2)
with col1:
fig = _render_wordcloud(analysis.get("positive_words", []), "Positive Language", "Greens")
if fig:
st.pyplot(fig)
with col2:
fig = _render_wordcloud(analysis.get("negative_words", []), "Negative / Hedging Language", "Reds")
if fig:
st.pyplot(fig)
# Divergence Alerts
alerts = analysis.get("divergence_alerts", [])
if alerts:
st.markdown("---")
st.markdown("#### ๐จ Divergence Alerts")
for alert in alerts:
st.error(f"โ ๏ธ {alert}")
# Between the Lines
st.markdown("---")
with st.expander("๐ฎ What Management Is Really Saying", expanded=True):
st.markdown(analysis.get("between_the_lines", "No analysis available."))
# PDF Export
st.markdown("---")
if st.button("๐ฅ Download Sentiment Report as PDF", key="es_pdf"):
from features.utils import export_to_pdf
sections = [
{"title": f"Earnings Sentiment: {ticker_display}", "body": f"Management: {mgmt.get('label', 'N/A')} ({mgmt.get('score', 0):.2f})\nQ&A: {qa.get('label', 'N/A')} ({qa.get('score', 0):.2f})"},
{"title": "Key Themes", "body": ", ".join(themes)},
{"title": "Between the Lines", "body": analysis.get("between_the_lines", "")},
]
pdf_bytes = export_to_pdf(sections, f"{ticker_display}_sentiment.pdf")
st.download_button("โฌ๏ธ Download PDF", data=pdf_bytes,
file_name=f"{ticker_display}_Sentiment_Report.pdf",
mime="application/pdf", key="es_pdf_dl")
|