NewsChecker

• 

  Original planning for our project

• 

  Transcription jobs in AWS

• 

  AWS shell output after successfully transcripting video

Inspiration

On the internet, misinformation is present in many sources including articles and videos. We wanted to give users the ability to look at articles and video transcriptions and see if they are misleading.

What it does

Our code is composed of two different programs. In aws.py, users can input the filename of a .mp4 file and receive the full transcript of a video. They can then copy and paste this transcript and classify it using the machine learning algorithm or directly provide a .txt file with the transcript to the algorithm.

How we built it

We built the transcription part in AWS. In the program, we use boto3 to call an AWS bucket that takes a video file as an input. Whenever the bucket receives an input, it causes a trigger event which calls an AWS Lambda function. The Lambda function takes in the input, which is a video, and transcribes the video and returns a transcription to another S3 bucket. Meanwhile, in the program, we keep on querying the other S3 bucket waiting for a response. Once we get a response, we process the response (which is a .json file) and print out the video transcript.

Challenges we ran into

At around 3:00 am, all of our individual parts were working well but our parts couldn't combine to work well. We couldn't get file uploads to work well with our UI, which we were implementing through flask. After spending hours working on it to no avail, we decided to just submit our individual components in python files so that users can just download our python files and run our programs.

Accomplishments that we're proud of

We are proud of creating a neural network that recognizes the difference between real and misleading news. We are also proud of using AWS to transcribe videos, as that took a lot of steps. We are also proud of persevering and finding solutions to problems even when it seemed all hope was lost.

What we learned

We learned how to use AWS transcribe and learned how to create a neural network to classify text. We also learned that front-end coding isn't simple and that pieces of programs that work separately won't work together combined.

What's next for NewsChecker

In the future, we would implement a better UI for NewsChecker so that it is easier for users to get video transcripts and classify articles.

Built With

• amazon-web-services
• awstranscribe
• kaggle
• python
• s3
• tensorflow