- Aim : Doing a sentiment map for Switzerland.
We have at our disposal :
-
Large twitter data (geolocated)
-
Instagram images (geolocated)
- Analysis :
-
Study our dataset
-
Perform a sentiment analysis using convolutional / recurrent deep neural networks for Twitter data (also NLP methods like word embedding)
-
Perform sentiment analysis using convolutional deep networks on instagram images: the advantage of convolutional neural network is that we don’t need to preprocess the image and extract relevant features we can feed the whole image to the model and optimize the parameters
-
For the analysis we will be using Keras
- Viz :
-
Building a sentiment map
-
You can find detailed explanations on the readme files in the folders : Tweet-Classification and Instagram-Classification
-
You can also find our poster presentation pdf under poster.pdf
- Yassine Benyahia : yassine.benyahia@epfl.ch
- Mohammed Hamza Sayah : mohammed.sayah@epfl.ch
- Juraj Korcek : juraj.korcek@epfl.ch
- Félix Böhm : felix.bohm@epfl.ch