Recent events such as the Cambridge Analytica scandal have prompted a heightened focus on personal data and its use by tech companies. However, most users are still faced with daunting service agreements designed to discourage in-depth reviews of the rights they are forfeiting. Legislation such as GDPR is a step in the right direction, but our team felt that technology should also be leveraged to empower users and take back control of our digital lives.
What it does
When faced with an extensive terms and agreements page, users can simply paste the URL into our custom website. Using Microsoft Cognitive Services, the document is analysed and made available for users to query via the Privacy Expert Alexa skill. Example questions include:
- Find the terms and conditions
- Tell me about subscriptions
- Give me details about content availability
How we built it
The custom website was built using Flask and accepts a URL input which then is sent to a Python backend script. The script makes an API call to QnA Maker on Microsoft Cognitive Services for extracting and generating Q&A pairs from the target URL. This result is then downloaded and processed using a DeepAI text summarisation API, before being uploaded back to a Cognitive Services knowledge base.
The Alexa skill uses an AWS Lambda function which processes the QuestionIntent and identifies the Q&A pair most relevant to the query.
Challenges we ran into
It was discovered that the QnA Maker output was not robust to varying terms and conditions formats on different websites; we attempted to address this issue with post-processing and format standardisation with some improvement seen. After discussing with Microsoft representatives, it was discovered that the URL setting for the QnA Maker could not be updated via http requests. Possible workarounds include the deletion and creation of a new knowledge base structure for each new target website; this will be addressed in the scalability process.
Accomplishments that we're proud of
It was our first time developing using Alexa, AWS, Azure, and QnA Maker; quite some time was spent on familiarising ourselves with the processes but we are glad that our final project was able to incorporate the functionality of all services and leverage the advantages of each.
We also genuinely believe that products like these will have a positive impact on the current discussions surrounding consumer data and make user rights more accessible to the public.
What we learned
We learned how to design an enjoyable Alexa voice experience that can handle a wide range of queries. We were also able to better utilise cloud computing resources, such as AWS and Azure, than in our previous projects. We designed the integration between APIs with future scalability in mind, and discovered the importance of this approach when reviewing our next steps for this project.
What's next for Privacy Expert
We hope to improve the robustness and scalability of the API services by provisioning more knowledge bases and creating a more comprehensive Python script for processing output text. We also look to build a database of the most commonly encountered service agreements (e.g. Facebook, Google, Apple...etc) such that users can immediately query the Alexa skill without additional setup. The final step would be to publish the Alexa skill and make it available to the users that need it.