Natural Language Processing Text Analysis Prompt

Analyze unstructured text data using NLP techniques including sentiment analysis, topic modeling, entity extraction, and text classification to uncover actionable insights from customer feedback, reviews, or survey responses.

Prompt Template

You are an NLP data scientist specializing in extracting business insights from unstructured text data. Help me analyze the following text dataset and build an actionable insights report.

**Data Source:** [e.g., customer support tickets, app store reviews, NPS survey responses, social media mentions, product feedback, employee surveys]
**Volume:** [approximate number of text entries — e.g., 500 reviews, 2,000 support tickets]
**Time Period:** [date range of the data]
**Business Context:** [what you're trying to learn — e.g., why churn increased, what features customers want, brand perception shift]
**Current Analysis:** [what you've done so far — e.g., read a few manually, have star ratings, nothing]

Here is a representative sample of the text data (provide 15-30 examples):
```
[Paste sample entries here]
```

Please perform the following analyses:

1. **Sentiment Analysis**
   - Classify each entry as Positive / Neutral / Negative with confidence score
   - Calculate overall sentiment distribution
   - Identify sentiment trends (if dates provided)
   - Highlight entries with mixed sentiment (positive about one thing, negative about another)

2. **Topic Modeling**
   - Extract the top 8-12 topics/themes from the data
   - For each topic: label, description, frequency (% of entries), representative quotes
   - Create a topic × sentiment cross-tabulation (which topics are positive vs. negative)

3. **Named Entity Extraction**
   - Products/features mentioned
   - Competitor mentions
   - People/team mentions
   - Specific metrics or numbers referenced

4. **Keyword & Phrase Analysis**
   - Top 20 most frequent meaningful phrases (excluding stop words)
   - Emerging phrases (increasing in frequency over time)
   - Co-occurrence analysis: which topics appear together

5. **Actionable Insights Report**
   - Top 5 findings with supporting evidence (quotes + data)
   - Recommended actions for each finding
   - Priority ranking by business impact

6. **Analysis Methodology** — Explain the approach so I can repeat this analysis on new data:
   - Python code snippets for automating the analysis
   - Tool recommendations (open source + commercial)
   - How to set up ongoing monitoring

Example Output

# NLP Text Analysis Report: App Store Reviews (Jan-Mar 2026)

**Dataset:** 847 reviews from iOS App Store | **Product:** TaskFlow mobile app

1. Sentiment Analysis

Overall Distribution

| Sentiment | Count | Percentage |

|-----------|-------|------------|

| Positive | 412 | 48.6% |

| Neutral | 198 | 23.4% |

| Negative | 237 | 28.0% |

Sentiment Trend (Monthly)

|-------|----------|----------|---------------|

| January | 54% | 22% | +32 |

| February | 48% | 28% | +20 |

| March | 41% | 35% | +6 ⚠️ |

**⚠️ Alert:** Net sentiment dropped 26 points in 3 months. March negative reviews are dominated by the "sync issues" topic (see below).

Mixed Sentiment Examples

> "Love the UI redesign, it's gorgeous. But since the update, my tasks keep disappearing when I switch devices. I've lost data twice." — ★★★ review

**Positive:** UI redesign | **Negative:** Data sync/loss

2. Topic Modeling — 10 Topics Extracted

|---|-------|-----------|-----------|---------------------|

Topic × Sentiment Matrix

Positive Neutral Negative

Sync/data loss ████░░░░░░░░░░░░░░░░ 11% 89%

UI/design ████████████████████ 92% 3%

Performance ██████░░░░░░░░░░░░░░ 22% 71%

Calendar request ████████░░░░░░░░░░░░ 25% 10%

Pricing ████░░░░░░░░░░░░░░░░ 15% 78%

Collaboration ██████████████░░░░░░ 68% 12%

Notifications ███░░░░░░░░░░░░░░░░░ 10% 82%

3. Named Entity Extraction

**Competitors mentioned:** Todoist (34×), Notion (28×), Asana (19×), Things 3 (12×), TickTick (8×)

**Features mentioned:** subtasks (67×), calendar view (54×), widgets (41×), dark mode (23×), recurring tasks (19×)

**Platforms:** iPhone (156×), iPad (43×), Mac (38×), Android (12× — "when is Android coming?")

5. Actionable Insights Report

Insight 1: Sync reliability is destroying your ratings 🔴

**Evidence:** 18.2% of all reviews mention sync issues, 89% of those are negative. March saw a 3× increase in sync complaints coinciding with v3.2 release.

**Impact:** Estimated 0.4-star rating drop attributable to sync issues alone.

**Action:** Investigate v3.2 sync architecture changes. Consider rolling back or hotfix within 1 week.

**Priority:** P0 — Critical

Insight 2: Calendar view is the #1 feature request

**Evidence:** 54 explicit mentions of "calendar" or "calendar view" across 847 reviews. Several 4-star reviews say "would be 5 stars with calendar view."

**Impact:** Potential to convert 30-50 reviews from 4★ to 5★.

**Action:** Prioritize calendar view in Q2 roadmap.

**Priority:** P1 — High

6. Methodology & Automation

# Quick sentiment analysis with TextBlob

from textblob import TextBlob

import pandas as pd

df = pd.read_csv('reviews.csv')

df['sentiment'] = df['text'].apply(lambda x: TextBlob(x).sentiment.polarity)

df['sentiment_label'] = df['sentiment'].apply(

lambda x: 'positive' if x > 0.1 else ('negative' if x < -0.1 else 'neutral')

)

# Topic modeling with BERTopic

from bertopic import BERTopic

topic_model = BERTopic(nr_topics=12)

topics, probs = topic_model.fit_transform(df['text'].tolist())

topic_model.get_topic_info()

Recommended tools:

- Open source: BERTopic (topic modeling), VADER/TextBlob (sentiment), spaCy (entity extraction)

- Commercial: MonkeyLearn, Qualtrics Text iQ, Thematic.io

- LLM-powered: Use Claude/GPT-4 for nuanced analysis on smaller datasets (<1,000 entries)

Tips for Best Results

💡Start with a representative sample of 50-100 entries to develop your taxonomy before running analysis on the full dataset. Garbage-in, garbage-out applies to NLP even more than structured data.
💡Sentiment analysis tools miss sarcasm and context. Always manually validate a random 10% sample — if accuracy is below 80%, switch to LLM-based analysis or fine-tune your model.
💡The most valuable insight from text analysis is usually the TOPICS, not the sentiment. Knowing that 18% of feedback mentions 'sync issues' is more actionable than knowing 28% is negative.
💡Set up automated weekly analysis (even a simple keyword tracker) to detect emerging issues before they become crises. A sudden spike in 'crash' or 'data loss' mentions needs immediate attention.

Try it with

ChatGPT Claude Gemini

Related Prompts

Data Analysis

Dataset Summary and Insights

Paste or describe a dataset and get an instant summary of key statistics, patterns, anomalies, and actionable insights.

ChatGPTClaudeGemini

Data Analysis

SQL Query Writer for Business Reports

Generate SQL queries for common business reporting needs — revenue trends, cohort analysis, funnel metrics, and more.

ChatGPTClaude

Data Analysis

Dashboard KPI Definition Framework

Define the right KPIs for your business dashboard with clear formulas, targets, and data sources.

ChatGPTClaudeGemini