E-Vroom

AI News - Full Stack

Comment

What inspired us to create E-Vroomers News?

BurdaForward’s mission to spread news about sustainability and electric vehicles inspired us. With E-Vroomers News we target any people who is interested in electric vehicles and sustainability. Whether it’s curiosity, whether it’s a school assignment, whether it’s understanding the technology before making a purchase we aim to provide useful content. We are a team of Computer Science students who are AI and web-development enthusiasts, and we found the challenge of “AI-Driven News Generation About electric vehicles” the most exciting.

How can you use E-Vroomers News?

  • How can you use E-Vroomers as a reader?

Just visit our WordPress website and click on any article of your interest to read it. You’ll see the category of each article on the website, and you can navigate to other articles after finishing one.

  • How can you use E-Vroomers as a content creator?

On our GitHub we provide a JSON file of automatically collected and filtered article titles and short descriptions. For generating and uploading the content of new articles we used ChatGPT4.0 and a WordPress plugin. You can create content by creating a ‘new post’ and on the side bar using our plugin implementation for uploading the JSON file and selecting a topic from the pop-up list, then clicking on the ‘enhance’. You can generate a picture related to the new article if you prompt our plugin with the title of the article.

How did we build it and how does it work?

Our project consists of multiple components.

  1. We wrote a web scraping script in Python utilizing the BeautifulSoup library to gather information about the news in RSS. Then we used filters to narrow down the information to the relevant topics: e-vehicles.
  2. We feed the relevant information in the form of title and short description pairs to a HuggingFace model for sentimental analysis. This way we can add a sentiment to the title and the short description then store them together in a JSON.
  3. For generating content, we use ChatGPT4.0. We determine the topic of the article based on input the JSON file. Then we use prompts which we crated based on Google Search Central Guidelines to ensure that we not only have a relevant article but also a good quality one.
  4. We use Dall-E for generating an image which is relevant for the article. Dall-E can be prompted with the title of the generated article.

We ran into a few challenges

The biggest challenge for us was to integrate the components together and create a fully automatized workflow. Most of the transitions between the automatic content creation are automatized, but uploading the JSON for the topic, hitting a few buttons and optionally reformatting the text is still the task of the content creator person. Originally, we wanted to use a database in which we can manage the JSON files and the content of the new articles, but we didn’t have time implementing that. At the beginning of the project we encountered smaller obstacles, like the plugin was not taking any effect, but we successfully overcame these ones.

Accomplishments that we're proud of

In general, we are proud of the project which we created throughout the weekend. It has a lot of areas for improvement, but essentially, it’s a working application. We have never done web scraping before, so implementing it was one of the achievements that we were very happy about. Also, we have never deployed AI in web-development, so it was another happy moment when the plugins for WordPress finally worked.

We learned a lot

We learned a lot about content generation with generative AI and deploying it in a web site. Starting with the brainstorming, having snack breaks, coding, getting more snack, doing more coding, debugging, trying to put components together and now creating our presentation we’ve been through a journey which was fun and gave us useful work experience for the future.

What's next for E-Vroomers News?

We would like to fully automatize the process of content creation, add a database to the workflow and create a more appealing user interface. If E-Vroomer News is succeeds we can try to use content generation in other news areas, like housing to help people learn more about this topic.

Built With

  • apache2(2.4.59)
  • chatgpt4.0
  • dall-e
  • debian(11.9)
  • mysql-client(8.0.37)
  • mysql-community-client(8.0.37)
  • mysql-community-server(8.0.37)
  • mysql-server(8.0.37)
  • php(8.1.28)
  • phpmyadmin
  • wordpress(6.5.3)
  • wp-cli(2.10.0)
Share this project:

Updates