Emotion Bot

Introduction

Background

Despite the fact that Facebook provides a translation feature by default, text translation is often insufficient due to different linguistic syntax and cultural context. Given the fact, we believe that intercultural interaction has been greatly hindered by mere translation.

Research Questions

What would help people to understand their foreign friends better when the translations fail?

Key Goals

Enhance intercultural communications through emotionn and sentiment analysis

Beyond Simple Translation

Our clients were Professor Sue Fussel and her Ph.D. student Hajin Lim. After discussing with Hajin and getting to know her visions on the project, we decided to do go with her prototype idea with an extra feature: sentiment analysis.

The goal of our team was to go beyond translation.

Sentiment Analysis Explorations

- Pointer and gauge to represent the scores -

Implementation

Feature

Sentiment Analysis

The meter chart we used for sentiment analysis showed the user to what extent the person was feeling negative or positive.

Emotion Analysis

5 different emotions: anger, disgust, fear, joy, and sadness. Each bar was colored differently and the higher the bar, the stronger the emotion.

User Case One

Original Text
Translation: in this case it is a correct translation
Sentiment: extremely negative
Emotion: mostly sad, but also angry, surprised, and joy

The news title can be translated as “3 Cultures that Japan Has to Abandon for Innovations to Take Place”.

User Case Two

Original Text
Translation: in this case it does not form a concrete sentence
Sentiment: very positive
Emotion: pure joy

Even though the translation does not quite make sense, combining the feeling and the emotion, users can interpret the meaning behind the post.

Design Decisions

Gauge instead of bar chart

It's hard to show neutral sentiment analysis score with bar chart.
Without showing score, it’s hard for user to know the intensity of the post just by its color.
Gauge provides a gradient from red (which is the most negative) to blue (which is the most positive). Even without the actual score, it shows users whether the post is negative or positive, and how negative or positive it is.

Emojis instead of cartoon faces

Since the facial expression of emojis are more obvious than the cartoon faces, we decided to go with emojis.
However, some emojis might be interpreted differently.

Direct click instead of button

Direct click and button serve the same purpose: to help user get to the analysis of the post. We want to provide users a simpler experience, so we change the button, which requires two steps, to direct click, which requires only one step.

Chrome Extension for Intercultural Communications