Guess My Fruit (Alexa Skills Challenge)

Inspiration

The latest app, Guess My Fruit V1.0 developed by UofT Engineering Science students! All content including fruit facts, and any images associated with this skill, are original creations. This was inspired by Akinator, the game that attempts to determine which character the player has in mind by asking them a series of questions.

What it does

Guess My Fruit is a skill which enables Alexa to guess the fruit you're thinking about. Just think of a fruit, and let Alexa know you're ready. She will ask you a few questions and then guess which fruit you had in mind! There are no accounts, no sign-ups, and no collection of any private information. It is a simple skill which enables kids to engage with a fun game while learning about the wonderful world of fruits!

How we built it

The front-end interface of the multi-turn dialog was created using the Alexa Skill Kit. Using various sentence structures, sample “utterances” between a player and the game were synthesized, leaving slots for the utterances that were registered as variables to the program. The logic of the conversation was assigned using an AWS Lambda instance. The program uses an algorithm that combines certain heuristics to determine the order in which questions should be asked to minimize the number of queries. Specifically, the algorithm references a dataset comprised of a list of fruits and a list of their attributes. To formulate each guess, the algorithm selects the attributes that have the most variability of entries. Every time a response is given by the user, the algorithm eliminates the possible fruits until there is only one left.

Challenges we ran into

As it was our first time developing for Alexa and using the Alexa Skills Kit, we initially had slight difficulties familiarizing ourselves with using the platform. The bulk of the challenge was to apply common sense in a way that would be entertaining and educational for children aged 7 to 13 and properly adapt for the audience. The most difficult task was figuring out how to prompt users for certain slots without making the slots “required”; thus, ultimately guessing a fruit efficiently and without the need to ask 15 questions each time (the full classification of fruits).

Accomplishments that we're proud of

We devised an algorithm to efficiently determine the average "shortest path" to reach the final answer of a guess trial, and made the database of fruits and facts easily scale-able. Our first idea was to use a giant "if" ladder, but thank god we didn't! Learning the platform in a few short days and putting the project together with a team has been great, and we look forward to our next experience!

What's next for Guess My Fruit

Guess My Fruit has a lot of potential for expansion. We plan to expand our database of fruits (which currently contains 28 fruits, which is plenty, but there are still more fruits out there!), and add a larger variety of facts for each fruit (we currently have 2-3 interesting facts for each fruit). A long term goal is to program the inverse of guess my fruit, and allow users to guess a fruit which alexa is thinking about by posing questions! This makes the skill even more fun, engaging, and educational for kids.

Built With

• alexa-skills-kit
• amazon-alexa
• amazon-web-services
• aws-lambda
• node.js