Logical Mixologist

Inspiration

While thinking about how Answer-Set Programming can be applied to real world explainable/interpretable AI systems, we struck upon the idea of a drink recommendation system.

What it does

Logical Mixologist takes in the user's preferences and combines them with internal knowledge to recommend suitable drinks. s(CASP) is able to provide the chain of reasoning - the internal model that led to the answer, ensuring recommendations are transparent.

How we built it

The core of the system uses s(CASP) to reason about what drinks to recommend to a user based on set preferences. We also built tooling using python to convert the database and user-preferences into s(CASP) queries. This freely-licensed dataset of drinks provided the knowledge for our system.

Challenges we ran into

We struggled with much of the tooling around s(CASP). Some of our files did not fit s(CASP)'s top-down evaluation model well (taking over 9 minutes to run), so we had to modify our process.

Accomplishments that we're proud of

We are proud of creating a prototype to showcase the power of s(CASP) to do logical inference.

What we learned

• Working with s(CASP)
• Cleaning data sets

What's next for Logical Mixologist

In the future, this system could be expanded with a web front-end using flask or django to allow end users to better interact with the s(CASP)-based system. We might also explore creating a mobile app to reach more users.

Built With

• prolog
• python
• scasp