The sentiment behind reviews is a Web application that help users get intuition about the product they intend to buy from online stores.
Each Notebook has its own explanation of functions in more details.
- Scraping Notebook
- Cleaning Data Notebook
- Different features enginnering & Modeling Notebook
- web Application
- Project Report
- Project Video
Go to "Graduation App" folder then write interminal:
- pip install -r requirements.txt
Test Client Review, Retrive Product and predict Senitment of all reviews
| Model & Features Engineer | CountVectorizer | TF-IDF | Word2Vec |
|---|---|---|---|
| Logistic Regression | 89 | 90 | 83 |
| SVC Linear | 88 | 90 | 83 |
| MultinomialNB | 86 | 84.1 |
Structure of the Scrapping & cleaning:
- connect to cloud mongodb database: some variables defined in souq_configs file.
- handle some firefox preference and options that help us during process of scrapping
- others files like cleaning and features engineering work on these scraped data, actually on reviews of products
- testing some of functions
- remove punctuations of all reviews
- remove stop words
- spell correction
- Tokenization
- Lemmatization
- Binary BOW
- CountVectorizer BOW
- TF-IDF BOW
- Word2Vec
- LogisticRegression
- SVC Linear
- MultinomialNB naive_bayes
- Function to compine the returned data frames to one data frame
- Another Function to shuffle the reviews of the returned combined data frame**
- Another Function to convert the data frame to csv file
- handle classifed reviews
After what we have discussed in all of these processes we end up with what the user can see now. The Sentiment behind reviews is designed to get products from online stores like Souq & Jumia to make analytical reviews of these products and predict the Sentiment of each review on the product (Positive Or Negative), besides a pie chart that represents a quick overview of the ratio between positive and negative.