Team Platypus

This is our solution for the Data and AI Hackathon, made with Azure Cognitive Services.

Comment

Note: The repository we added here contains links to other repositories/resources in README.md

Inspiration

We have been inspired by the power of AI and how we can use it to find detailed information on how people around the world talk about a brand. Is our brand ever mentioned in Spanish? Do people speak positively about our brand? Is the brand name often mentioned along with other brands or specific people? We want to use AI to find details on how people speak about a brand, and with that, get insights into where the company should focus their efforts in improving the brand.

What it does

Our interactive Adobe XD prototype displays big parts of the app we want to develop. The main idea in this app is that the company will run an analysis, which is based on search terms, platform (e.g. specifically facebook, or Bing for a wider analysis), and a date span. Depending on those criteria, the Azure services will create an analysis that can tell the company about

  • how many times, the search terms have been mentioned within the date span;
  • which key words has it been mentioned alongside;
  • are the texts, where the keywords are used, generally positive or negative;
  • and much more.

The search terms can obviously be the brand name; however, it can also be a specific product, the branch or something else, that can be useful to analyze. Our prototype also contains a comparison feature, where a company can compare the results of their own brand and a competing brand, the results of two products etc.

How we built it

To get a functional prototype of parts of this project, we have chosen a path where we use Azure Cognitive Services to:

  • web-crawl with Bing;
  • analyze the texts we find with AI text analytics.

The Cognitive Services Text Analytics is the main part of our project. This is the part that analyzes the texts and turns it into useful data.

For a simple demonstration of displaying the data, we have created a small Vue SPA with a table with mock data. This displays how easy and useful it can be to compare the results from our analytics.

Challenges we ran into

We have had some issues figuring out exactly what services we should use, especially for web-crawling. At first, we wanted to focus specifically on analyzing the brand on a specific platform, e.g. facebook, but we didn’t find the proper way to do this. Instead, we went with using the Bing search machine for finding brand mentions. This has the advantage that it is very wide and can give a general picture of how the brand is depicted online; however, it is more difficult to keep track of dates and changes in how people speak of the brand, and whether people speak differently at different platforms. Therefore, a future addition to the app will be the possibility to analyze specific platforms, like defined in our prototype.

The future of the app

There are of course a lot of things that generally still need to be created and coded. But other than just continuing on our current path, there are also other things we’d like to add to our app.

  • We would need to spend a lot more time and resources on researching and defining proper categories, comparisons etc. to analyze. Our current XD prototype is built on loose ideas of what can be useful, and we have focused on making it visually attractive and easy to get quick overviews. It would, however, need more ways to analyze more focused and in detail.

  • As mentioned earlier, we want to be able to analyze more specifically on the different platforms. This would also include “News” as a general platform, to see how the media talks about products etc.

  • We would like to add image analysis. This would also be done with Azure Cognitive Services, which also includes Computer Vision for image analysis. Our initial idea for this would be to search for images that include a brand logo, and analyze the images for:

  • How often this occurs;

  • Which types of items typically appear in photos;

  • If other brands are in the photos;

  • If certain people are in photos;

  • If the photos are generally light or dark.

This can, just like text analysis, help the company see how their brand is perceived by the public, how it is actually used, and which areas they might want to target for promoting and branding.

Accomplishments that we're proud of

We’re proud of our design, which we believe both looks good and is intuitive to the user (the company making the analyses, e.g. IKEA, Spotify etc.).

We’re also proud that we got this far using the Azure features and connecting them. We wouldn’t have needed much more time to get some analysis results into the actual Vue front end, so we’d be able to display the data the way it’s supposed to be displayed to the user.

What we learned

We have generally all learned a lot during this project, seeing that we hadn’t worked much with Azure or AI before, and especially not in connection to other features. We all decided from the beginning that we wanted to use this opportunity to teach and learn, meaning that some of us have taken a leap of faith and started learning JavaScript, TypeScript, nestJS, upgraded to Vue 3 etc and more.

What's next for Team Platypus

We have had a lot of fun and great teamwork in this group, and we will definitely be open for another hackathon with this group. We have also had the perfect mix of different skills (design, front end, back end etc.), and we all now know some more people that we can go to for help if we need it.

Built With

Share this project:

Updates