Proposal

Problem Statement

Career development is one of the most significant goals of college students. More and more students are searching for internship, full-time or research opportunities through networking, career fairs, interviews, career-development conferences, etc. During these career-development activities, politeness is extremely important for showing your respect to the person you are talking to, demonstrating your ability of professional communication, leaving a better impression, and winning more chances of getting an offer. This project aims at improving college students’ level of politeness during professional communications by analyzing the level of politeness from the user’s tone as well as their speaking rate and giving detailed suggestions on how to improve. To use this system, user can enter the website, record a short audio of himself speaking, and view data analysis on his speaking sample. The site is supposed to collect and analyze the audio data. The analyzed results will include the level of their audio’s politeness (in percentage), the pitch and speed of their words, and how they could improve their level of politeness. Despite that we are focusing more on the career development aspect in this project, our design can also be used to improve politeness in daily life settings so that people’s communication efficiency can be boosted.

NeedFinding

Interview with a faculty, a Career Center advisor, and a speaking fellow: The goal of the interview is to find out the necessity of improving politeness as well as understanding what features could be important in determining level of politeness in professional communications.

Potential Interview Questions

1. Do you think that college students often seem polite enough in professional communications?
2. If yes, how do you see that? If no, where do you see signs of impoliteness?
3. Do you think speaking tone matters in the level of politeness?
4. What kinds of tone could show a sense of politeness in professional communications?
5. What suggestion would you give to help students improve their level of politeness in professional communications, from the aspect of speaking tone?
6. Do you think a system giving suggestion on speaking tone could be helpful in improving level of politeness?
7. Do you think politeness is required for professional conversation? To what extent?

Survey: Survey will be distributed to students. The goal is to discover whether improving level of politeness in professional communication is needed among college students. The survey would be distributed through various social platforms like Facebook. We would try to include students from different backgrounds.

Potential Survey Questions

1. How often do you participate in professional communication? (such as career affair, networking events, internship interview, etc.)

a. 1-2 times per month

b. 1-2 times per semester

c. 1-2 times per year

d. Never

2. I struggle with my politeness skill in a professional communication.

a. Often

b. Sometimes

c. Hardly ever

d. Never

3. I worry that I did not speak politely enough after I finished an interview.

a. Often

b. Sometimes

c. Hardly ever

d. Never

4. Which of the following could be a sign of impoliteness in professional communication?

a. Volume too low or too high

b. Speed too fast or too slow

5. Which of the following aspects would you like to practice more?

a. Volume

b. Speed

6. Would you be interested in a website that could help adjust your tone and improve your level of politeness? (from 1 to 5)

a. 1(not interested at all) b. 2 c. 3 d. 4 e. 5(strongly interested)

Prototyping

The prototyping portion would include two parts: low fidelity wireframe using Axure and high fidelity prototype (static pages) using Sketch and InVision. After the static pages are created, there will be a clickable prototype made using InVision in order to roughly test the logic behind the entire structure. The wireframe would show basic structure of the application while the high fidelity would focus more on details such as background colors, typography, shapes of elements, as well as some basic interactions between users and webpages. The prototype will try to demonstrate all the features as well as functionalities (breadth and depth) in our design, but the cases demonstrated will be mostly fixed because the actual product requires more back-end support. After the prototype is finished, we will test it within our group and recruit five other users via Facebook to evaluate the interactions to make sure the design makes sense. In order to test the prototype, we are going to test it among the group as well as 5 other students, who will be given various tasks (e.g., “suppose you have already finished the evaluate and you just want to see your results history, show me your steps”, “you are a new user and you want to do an evaluation, show me the steps”), and the entire process will be observed to see how often they find the solution without clicking the wrong button. The prototype will also be tested by the Career Center advisors as well as speaking fellows, just to see if other features can be added.

Implementation

The system we’re proposing has four key elements for us to implement: a view, a controller, a model, and a dataset.

We intend to collect a dataset consisting to train and evaluate our model. This data set will be based on the dataset collected by Chiharu Tsurutani and Shuju Shi. Their dataset consists of 18 Japanese sentences read by a total of 29 people, including native and non-native speakers of both genders. Our dataset will consist of English translations of the sentences used by Tsurutani and Shi. We will record adults of both genders between the ages of 18 and 25 to create this dataset. We will include both native and non-native speakers in order to maximize the likelihood that our system works for all English speakers. In addition to wav files, the dataset will contain plaintext files representing preprocessed versions of each recording.

The purpose of the view is to present data to and collect data from the user. We have decided to make the view a webapp to maximize portability. We will rely on bootstrap to implement much of the functionality of the view. The layout itself we be determined through needfinding and prototyping.

The purpose of the controller is to facilitate the communication between the view and the model. For our project, we intend to implement the controller as a server written in python. The server will accept user input from the view and feed it to the model, then take the information from the model and present it to the user as part of the view. Initially we will look towards Amazon Web Services as a host for the server. If that doesn’t work out then we will looks towards hosting the server locally.

The model is the process that scores the politeness of a given audio sample. We believe we can use a linear classifier to measure the politeness of a spoken phrase with sufficient accuracy. We intend to implement the model in Python. The model will reside on the same machine as the controller, easing communications between them.

User Evaluation

Our hypothesis for evaluation would be using tone to evaluation politeness is more accurate than using text. Thus, our control condition would be the way of interpreting speech (i.e., in text or evaluating pitch/tone). In order to test the result, a within subjects experiment will be conducted, with potentially 12 participants (6 female, 6 male, with 2 faculties and 10 students ranging from different school year). All participants would be recruited via Facebook and/or personal interaction. Participants would be asked to test both the Polite-o-meter that works on text and our design (the one that works on tone/pitch). We will evaluate our proposed solution based on efficiency (how fast users can get the results and how accurate and detailed the feedback is) as well as user feedbacks. More specifically, three things will be evaluated: model accuracy, user perception, and system usability. In order to test user perception, users would be asked to fill out a short survey about how confident they are about their level of politeness (e.g., tone, speed, etc.) before and after they use the system. They will also be asked to evaluate the effectiveness of the system after they use the system. In order to test accuracy, we will observe their entire evaluation process and detect the error rate. For system usability, we will ask users to fill out the standardized SUS test questionnaire in order to get a sense of the entire user experience.

Timeline and Deliverable

Alternative Solution

The original plan is giving users the percentage of politeness with the detail information about her/his voices(ex. Pitch, tone).If we can analyze the voice and help people to have more politeness voice, it would help them to be more socialized person. If it hard to approach the final goal, we would’ve just decide whether the voice is polite or not. It will still help users to know about politeness of their voices. We also can evaluate the politeness using the text.

Bio

Concept Video