Check It!

Have you ever been overwhelmed by wondering about which side articles are favoring? Then Check It! is for you! You can easily input details about your article, and it will quickly advise you whether your article is politically right or left. Our app uses a feed forward neural network to check through the article content to determine the article’s standing. This has been vigorously tested using many different news stations with reported standings to make the results more accurate. As we were making this app, we faced many hurdles such as upgrading and downgrading software to solve compatibility issues, learning new APIs which would work with our code, and many other errors that were caused by tensorflow’s lack of support. To be specific, the compatibility issues faced, were due to the tensorflow module in tensorflow v2 not having the attribute of sessions which was crucial for keeping the data of the tf.graph. As such, we found great difficulty in testing the cosine distance algorithm between two word vectors leading to problems with us attempting to classify if a word was pro-left or pro-right. The neural network we did was one which had two dense layers, each with 100 nodes with relu activation functions. Between each of these dense layer nodes, we have a dropout layer each to avoid overfitting of the data set. At the end of this, we plan to have a softmax function and two output nodes with one representing politically right, and the other representing politically left. Our proposed loss function was the cross entropy loss, and our optimizer was meant to be adam, an optimizer with a learning rate of 0.0006. Because of the setbacks stated before, we did not make a working demo. However, we believe that our project has the potential to have a positive impact on society, as it allows everyone to have a fair chance to know the political bias of articles they read.

Built With

• flask
• keras
• python
• tensorflow