Python Sentiment Analysis on Reviews -- 2
Budget: $250 – $750 USD
I have a collection of customer reviews that must be transformed into clear, data-driven insight. The raw files might arrive as CSV, Excel, or even straight from a database—the format is flexible—so the first task is to clean and standardise whatever I supply. That means handling empty rows, removing noise (HTML tags, punctuation, stop-words, etc.) and normalising the text.
Once the data is tidy I need robust sentiment classification. Please build the pipeline in Python, making sensible use of NLTK and/or TextBlob for tokenisation, lemmatisation and polarity scoring. The reviews span more than one language, but the immediate focus is on the English subset; your code should therefore detect language and process only English for this milestone while staying extensible for future Spanish or French additions.
I also want the story behind the numbers. Generate clear visualisations—think sentiment distribution charts, time-series trends, maybe a word cloud of highly polar terms—so stakeholders can grasp the overall mood at a glance.
Deliverables:
• Cleaned dataset with an added sentiment label
• Well-commented Python script or Jupyter notebook
• At least three informative visualisations (PNG or embedded in the notebook)
• A concise summary report explaining methodology, key findings, limitations, and next steps
Everything should run end-to-end on standard Python 3.x with common libraries (pandas, matplotlib/Seaborn, NLTK, TextBlob). If additional packages are essential, note them in a requirements.txt.
That’s the full scope—if the workflow sounds straightforward to you, I’m ready to get started.
Once the data is tidy I need robust sentiment classification. Please build the pipeline in Python, making sensible use of NLTK and/or TextBlob for tokenisation, lemmatisation and polarity scoring. The reviews span more than one language, but the immediate focus is on the English subset; your code should therefore detect language and process only English for this milestone while staying extensible for future Spanish or French additions.
I also want the story behind the numbers. Generate clear visualisations—think sentiment distribution charts, time-series trends, maybe a word cloud of highly polar terms—so stakeholders can grasp the overall mood at a glance.
Deliverables:
• Cleaned dataset with an added sentiment label
• Well-commented Python script or Jupyter notebook
• At least three informative visualisations (PNG or embedded in the notebook)
• A concise summary report explaining methodology, key findings, limitations, and next steps
Everything should run end-to-end on standard Python 3.x with common libraries (pandas, matplotlib/Seaborn, NLTK, TextBlob). If additional packages are essential, note them in a requirements.txt.
That’s the full scope—if the workflow sounds straightforward to you, I’m ready to get started.