My Planet

Inspiration

In China, people are unwilling to go to the psychological doctor. According to the investigation conducted by the People’s Daily, there are more than 100 million adults suffering from psychological problems and only one third of them received psychological treatment. The main reason behind this phenomenon is that people are reluctant and unwilling to go to psychological counselors. We want to build a platform to encourage people to record their everyday feelings and provide people who are likely to suffer form mental disease with psychological consultation if they want.

What it does

We built a platform for users to record their mental states as well as find potential patients with mental illness. Users need to input a text or voice describing their mental states into the platform every day. It can be the mood of the day, or a reflection of what happened. The software will use the machine learning algorithm to score the user's mental states, and send the records of users who are most likely to be liable to depression, anxiety, etc. to the professional psychiatrist. Based on these records, the psychologist judges whether a patient needs psychotherapy. Once a professional doctor thinks psychological intervention is necessary, he will inform the user of the help that he can provide. After receiving the doctor's notice, the user can decide whether he or she wants to establish a further relationship with the doctor, sends his or her own information, as well as accepts the psychotherapy. It is worth noting that these records about users are secret and cannot be seen by others. The professional psychiatrist can only see the words you write before you establish a doctor-patient relationship with him, without knowing your personal information.

How I built it

Using our app interface, users can enter their own words and save them. In the background, we use pytorch to create a recurrent neural network for semantic analysis of text. A feature vector extracted from the middle layer of RNN is inputted into a GAN for generating some pictures used to build a feedback image. The classification result is referred when sending notifications to doctors for further treatment.

Challenges I ran into

Insuficient data to improve the accuracy It’s unknown whether our UI satisfy users’ needs We provide users with a platform to express themselves and alleviate the potential threat for users’ stress and anxiety, thus providing a happier life for the users.

Accomplishments that I'm proud of

We build a recurrent neutron network to analyze users’ feelings and GAN to generate the items on the planet which accord to their recent feelings. Our RNN model has enough adaptivity to the stat-of-art algorithms to improve the performance of sentiment analysis. We have a great feedback system which can captivate our users.

What I learned

We find a effective app which can help people get better psychology helth should have a easy UI. It must be friendly and simple enough for the users to use at any time. Artificial intelligence such as neural network are effective in text detection

What's next for My Planet

We plan to provide better sentment detection service by improving our RNN model to the state of art. And we are going to take the planet as a API connecting various kinds of interesting applications. With more data acquired, we can provide more accurate sentiment consultion and provide patients with appropirate doctors or hospitals if necessary.

Built With

• python-rnn