SentiScore- for amazon products

In this project, I am upgrading the star rating of a product on e-commerce websites by combining the sentiment score of the reviews of the same product along with the star rating.

Comment

Inspiration

I have seen Amazon product reviews where customers' star rating doesn't seem to match with their text reviews. Hence I felt it is also important to convert text reviews into numeric data and provide a score that is a combination of star rating and text reviews.

What it does

The link of the product review is fed as input to the model and output is sentiScore of the product. It uses web-scraping and Modzy AI text summarization and Modzy AI sentiment analysis to produce the output.

what is sentiScore?: sentiScore is calculated by combining 75% of the average star rating and 25% of the average of the product review's sentiment.

How I built it

1.The first thing I had to do was web scraping Amazon product reviews. For this, I used beautifulSoup python package.

2.The second step was to pre-process the data scraped from reviews. For this, I used regex to remove numerical from the reviews and nltk's stop words to remove unwanted words from reviews. Then I stored the data into a dataframe using pandas.

3.Next, I created two functions. One for Modzy AI text summarization API and the other for Modzy AI sentiment analysis. Then stored the positive, negative, and neutral scores in individual lists.

4. As the final step, I combined 75% of the average star rating and 25% average sentiment score to get the final sentiScore.

Challenges I ran into

1.Without using Modzy AI text summarization I couldn't achieve the desired results.

2.Numericals were a problem until I used regex

3.I was new to web scraping. Hence had to face issues while scraping data and preprocessing them.

Accomplishments that I am proud of

I am proud of every small accomplishment I made in this project. Web scraping, integrating Modzy AI API's and generating the sentiScore. were all part of my accomplishment.

What I learned

I explored different use cases of Modzy AI's API service.

I learned a new skill: web scraping. I improved my coding knowledge. I learned to use APIs.

What's next for Amazon product reviews - sentiment analysis

1.I am going to improve the model's accuracy by using NLP BERT.

2.I wanted to create a web application for this project. Since I am not good at web development I have included this for future development.

3.In this user-friendly web application, people can enter the product name to view its sentiScore. This will also include features like comparing sentiScore of different brands of the same product so that people can choose the best brand for the product they wish to purchase.

Built With

Share this project:

Updates