truereviews.tech

Inspiration

Whenever we are looking to purchase a new product that may be use on a regular basis in our lives, we tend to pay close attention to this decision. As such, I find myself watching videos upon videos of product reviews. Whether it is a phone, a laptop, or a car. The idea is to get reviews from different sources so I can get an overall unbiased opinion and form an informed decision. This process, however, takes time.

What it does

We are building a website/tool that goes on the web, does a sentimental analysis of reviews from multiple different sources using our machine learning model, and returns an aggregated score for the product of interest. The beauty of this is, that since the data is sourced from multiple channels on YouTube, the website will always return truly unbiased information. If we were to run sentimental analysis on data from a single source, it would be prone to bias. For this case, our product of interest is the most used devices on the planet: Phones.

How we built it and Challenges

We accomplished this task by first getting the data from YouTube reviews captions using Google’s API and Python. Then, we modified the data for usability, I.E preprocessing. We then put the data in a CSV file for ease of access and efficiency in the short term because of lack of time and for efficiency. We then created our own machine learning model. The reason we didn’t use pre-established accessible models is because they are too exact, and our source, being YouTube video transcripts, is subject to regular speech patterns; as such, the analysis for each product was over-scoring. We wanted to be able to adjust the model according to our needs and home in on the ideas in the text we were looking for.

Accomplishments that we're proud of

We get an accurate sentimental analysis from video captions of video reviews base off of our own model, which we are proud of.

What we learned

How different aspects and areas of tech come together: IE Databases, machine learning, front and back end development, business, user experience etc.

What's next for truereviews.tech

Features: Sub-feature feedback sentimental analysis so we dont just get total rating of the product but also sub-fields, like camera quality and battery life

Business and monetization ideas:

We can run adds on the website We can have affiliate links to each of the product being shown for users to directly access and purchase the tool, we get a portion of the profits. Flat service purchase amount after a trial run for one product review.

Built With

• cockroachdb
• django
• domain.com
• machine-learning
• python