HALp Me Read
All our team members have previously conducted some form of research, and we are all painfully aware that most of our literature review time was spent sifting through - rather than carefully reading - papers.
This was mainly because it was difficult to quickly scan through and find papers that adequately pertained to our research topics. Thus, we resolved to build a tool that could quickly determine which papers held content that was important to us.
What It Does
HAL determines the three most important topics in a research paper's abstract using NLP
After processing each paper, our modified Alexa (dubbed "HAL" because of the pun potential) accepts two main inputs:
Get Recent [Topic]
Returns the most recent 15-50 publications that best fit the topic as determined by NLP
Check Files [Topic]
Scans through the user's personal library for publications that best fit the topic as determined by NLP
How We Built It
We used API.AI to create initial mockups of the chatbot flow. For our natural language processing, we employed the Google Cloud NLP API. For Alexa development, we worked primarily with Amazon Web Services.
Challenges We Ran Into
Initially, we thought API.AI could directly integrate with the Alexa Skills Kit, but the provided setup was unable to do so successfully. We were forced to use the API.AI as a template for hard-coded Alexa conversation paths. Later, the Google Cloud NLP API was a challenge to attempt integration with Amazon Web Services.
Accomplishments That We're Proud Of
We integrated powerful machine learning tools into our project, and gained valuable experience working with several robust development frameworks for the first time. In addition, we are proud to have worked together as a team where each member meaningfully contributed to the project.
What We Learned
Making frameworks work in tandem can be difficult.
What's Next for HALp Me Read
We aim to rank papers by their impact on the academic community by factors such as the reputation of their published journal and citation frequency.
Impact factor of paper so rank papers by reputation of journal, citation frequency, make it seamless like chatting with a colleague and has a grasp of concepts, hosting completely on Google Cloud, fact checking.