Storefronts and advertisers often have extreme amounts of data but are unable to extract meaningful use out of it. We were struck by the rich native APIs that NCR exposed as well as the lack of visual tools given to us to view it. This suggested that visualizing and analyzing the data, with the help of ML SaaS apis, could provide a better view into the data.
What it does
Our app periodically populates a MS Azure SQLSERVER with data scraped from the NCR api and applies deep and meaningful categorization on individual subpurchases on single items within a purchase. Our app extracts a highly detailed and non-generalizable name such as "Huggie's Rompers" and uses the Bing Search API to gain additional context. We then use the Azure Cognitive Services API to find key entities from snippets of additional context we were able to scrape.
How we built it
We split work into a front-end and back-end portion. We generally focused on creating mock data first to try to understand what format we would like the data to be in and worked off that to ensure the api did not diverge too much from the type of data our front-end app required.
Challenges we ran into
We realized hours in that SQLSERVER was very different from other SQL servers we previously had experience with. This lead to a harsher learning curve than expected earlier on and took up a lot of time. We were also having issues with implementing the ESRI-backed geospatial front-end using React. We eventually found key help from mentors who suggested helper methods for access to class types that we needed.
Accomplishments that we're proud of
We were able to create an app which responds relatively quickly and with rich support for additional querying. We are also proud of being able to recognize that we can extract additional contextual information about a product using Bing Search APIs and using
What we learned
We were able to learn new technologies and wire together completely disparate APIs to generate a functioning service. We also learned to use a complex multi-layer geospatial visualization software to display our results.
What's next for weekAtAGlanse
We want to be able to work to add additional filering on the data and to