ReportAnalysis

A powerful, batteries-included Python package for analyzing reports.

Drop in any report — as a text string, PDF, Word document (.docx), or URL — and instantly get sentiment analysis, readability scores, keywords, summaries, named entities, language detection, and much more. Ships with a full CLI and export to JSON, CSV, and HTML.

---

Features

Feature	Description
Sentiment Analysis	VADER + TextBlob ensemble with confidence scoring
Readability Scores	Flesch, Gunning Fog, SMOG, ARI — all computed offline
Keyword Extraction	TF-IDF keywords + RAKE multi-word keyphrases
Extractive Summary	Top N most informative sentences
Text Statistics	Word count, reading time, vocabulary richness, and more
Named Entity Recognition	People, Organizations, Locations via NLTK
Language Detection	Detects 50+ languages with ISO codes
Report Comparison	Cosine similarity score between two reports
Multi-format Loaders	Plain text, PDF, DOCX, and web URLs
CLI	analyze / compare / summarize subcommands
Export	JSON, CSV, and self-contained HTML reports

---

Installation

Minimal install (core only)

pip install ReportAnalysis

Full install (with PDF, DOCX, URL loaders and all analysis features)

pip install "ReportAnalysis[full]"

---

Quick Start

From a text string

from report_analysis import ReportAnalyzer

ra = ReportAnalyzer("The quarterly results exceeded all expectations. Revenue grew 30%.")
result = ra.analyze()
result.show()  # Prints a formatted report to the terminal

From a file

from report_analysis import ReportAnalyzer

# Supports .txt, .pdf, .docx
ra = ReportAnalyzer("annual_report.pdf")
result = ra.analyze()

print(result.sentiment.label)          # "positive"
print(result.readability.grade_level)  # "College"
print(result.keywords.top_keywords[:5])

result.export("analysis.html")  # Export as a standalone HTML report

From a URL

ra = ReportAnalyzer(url="https://example.com/annual-report")
result = ra.analyze()
result.export("results.json")

From a DOCX file

ra = ReportAnalyzer("report.docx")
result = ra.analyze()
print(result.summary.text)

Run only specific modules

result = ra.analyze(
    include=["sentiment", "readability", "keywords"],
    top_keywords=15,
    summary_sentences=5,
)

Compare two reports

ra1 = ReportAnalyzer("Q1 report text here...")
ra2 = ReportAnalyzer("Q2 report text here...")

comparison = ra1.compare_with(ra2)
print(f"Similarity: {comparison.similarity_score:.1%}")  # e.g. "72.3%"
print(comparison.similarity_label)                        # "Similar"
print("Common words:", comparison.common_words[:10])

Export results

result.export("analysis.json")  # Machine-readable JSON
result.export("analysis.csv")   # Spreadsheet-friendly CSV
result.export("analysis.html")  # Standalone HTML report

---

CLI Usage

# Analyze a file
report-analysis analyze report.pdf

# Analyze from a URL
report-analysis analyze --url https://example.com/annual-report

# Read from stdin
echo "Revenue increased by 30% this quarter." | report-analysis analyze -

# Run only specific modules
report-analysis analyze report.txt --include sentiment --include keywords

# Export results to HTML
report-analysis analyze report.pdf --export html --output results.html

# Compare two reports
report-analysis compare q1.pdf q2.pdf

# Summarize with 10 sentences
report-analysis summarize report.docx --sentences 10

# Show help
report-analysis --help
report-analysis analyze --help

---

API Reference

ReportAnalyzer(source="", *, url="")

Parameter	Type	Description
source	str	Raw text string, or a path to a .txt, .pdf, or .docx file
url	str	URL to fetch and analyze (keyword-only argument)

.analyze(include=None, summary_sentences=5, top_keywords=20)

Runs the analysis pipeline and returns an AnalysisResult object.

Parameter	Default	Description
include	None (all modules)	List of module names to run
summary_sentences	5	Number of sentences to include in the summary
top_keywords	20	Number of keywords to extract

Available modules: "stats", "language", "sentiment", "readability", "keywords", "summary", "entities"

AnalysisResult — Fields

Field	Type	Description
.stats	StatsResult	Word count, sentence count, reading time, vocabulary richness
.sentiment	SentimentResult	Label (positive/negative/neutral), compound score, confidence
.readability	ReadabilityResult	Flesch reading ease, Gunning Fog index, grade level
.keywords	KeywordsResult	TF-IDF scored keywords and RAKE keyphrases
.summary	SummaryResult	Extractive summary as sentence list
.entities	EntitiesResult	Named entities grouped by type
.language	LanguageResult	ISO language code and human-readable name

AnalysisResult — Methods

Method	Description
.show()	Print a rich formatted report to the terminal
.export(path)	Export to .json, .csv, or .html
.to_dict()	Return the full result as a Python dict

---

Result Details

Sentiment

result.sentiment.label             # "positive" | "negative" | "neutral"
result.sentiment.compound          # -1.0 to 1.0
result.sentiment.positive          # 0.0 to 1.0
result.sentiment.confidence        # "high" | "medium" | "low"
result.sentiment.textblob_polarity      # TextBlob polarity score
result.sentiment.textblob_subjectivity  # TextBlob subjectivity score

Readability

result.readability.flesch_reading_ease   # 0-100 (higher = easier to read)
result.readability.flesch_kincaid_grade  # US school grade level
result.readability.gunning_fog           # Years of education needed
result.readability.smog_index            # SMOG grade level
result.readability.reading_ease_label    # "Very Easy", "Standard", "Difficult", etc.
result.readability.grade_level           # "High School", "College", etc.

Keywords

result.keywords.tfidf_keywords   # [(word, score), ...]
result.keywords.rake_phrases     # [(phrase, score), ...]
result.keywords.top_keywords     # [word, ...] — plain list
result.keywords.top_phrases      # [phrase, ...] — plain list

Summary

result.summary.sentences         # ["sentence 1", "sentence 2", ...]
result.summary.text              # Joined summary as a single string
result.summary.reduction_ratio   # 0.0-1.0 (proportion of text removed)

Named Entities

result.entities.people           # ["Steve Jobs", ...]
result.entities.organizations    # ["Apple Inc.", ...]
result.entities.locations        # ["Cupertino", ...]
result.entities.entities         # {"PERSON": [...], "ORGANIZATION": [...], ...}

---

Dependencies

Installed with the core package:

• nltk — tokenization, VADER sentiment, named entity recognition
• click — CLI framework
• rich — terminal output formatting

Installed with [full] extras:

• textblob — secondary sentiment signal and subjectivity scoring
• scikit-learn — TF-IDF keyword extraction and cosine similarity
• rake-nltk — RAKE multi-word keyphrase extraction
• langdetect — language detection
• pdfplumber — PDF text extraction
• python-docx — Word document (.docx) loading
• requests and beautifulsoup4 — web page fetching and parsing

---

Running Tests

# Install development dependencies
pip install -e ".[dev]"

# Download required NLTK data
python -c "import nltk; nltk.download(['vader_lexicon', 'punkt', 'punkt_tab', 'averaged_perceptron_tagger', 'averaged_perceptron_tagger_eng', 'maxent_ne_chunker', 'words'])"

# Run the full test suite
pytest tests/ -v

---

Publishing to PyPI

pip install build twine
python -m build
twine upload dist/*

---

License

MIT License — see LICENSE for details.

Author

Al Mustafiz Bappy

• Website: almustafizbappy.zerodevs.com
• GitHub: @bappy-3
• PyPI: pypi.org/project/ReportAnalysis