The most crucial point of the project was to research characteristics of human natural language and suggest solutions that can reduce the gap between computer language and natural language for a smoother user experience.

‍

For example, "What should I eat today...." This sentence has two problems.

‍

First, it can be interpreted as "Recommend a menu to eat today" but, without a specific command like 'menu recommendation', the smart speaker cannot be smart.

‍

The second is that it is difficult to grasp 'what is meaning of today'.If these experiences were accumulated, it could be a bad user experience, like 'the speaker cannot understand well' or 'searching is faster than using a speaker'

Among the ranges of information that can be disclosed under the NDA contract, there are two solutions that I suggested

‍

1) Keep asking what is meaning of today to the user: to figure out what exactly the user wanted, keep asking questions using keywords related to 'today' such as schedule and weather

‍

2) Recommendations based on the user data such as anniversary, food the user cooked recently, trendy foods in social media. It helps to find the relevant answer even if users do not know what they want.

For making user scenarios, first I produced three different personas based on user data and analysed the conversation between the user and the speaker in the kitchen. To create an Information Architecture (IA), I  organised the necessary functions throughout the cooking process and designed screens of the whole process.

‍

‍

‍

‍

‍

‍