Our Inspiration for HOMA first stemmed from what can help make businesses more efficient. We wanted to incorporate mathematical calculations since every single business uses them. Furthermore, Many small businesses don’t have the resources to hire employees who can perform these tasks for them. HOMA can act as a substitute to those employees and help businesses perform the same tasks at a much lower cost. We landed on money calculations as our main topic of focus.
HOMA uses a speech to text algorithm to convert a live recorded audio of a word problem to a string. It then does the calculations in the back end to return the answer to said problems. As of now, HOMA can do maximum/minimum revenue and interest problems, but since it is just a prototype there will be more problems HOMA can solve in the future.
One challenge we ran into was trying to convert the speech to text using assembly AI. At first, we did not know how to use assembly AI proficient enough to record the users input through sound, with help provided with the MacHacks assembly AI sponsor we were able to overcome this hurdle and implement it efficiently into our code. Another problem we ran into was trying to assign what each number value in the converted string meant. The way we solved this was by finding the keys words around the number to see what the number represented in the problem.
Foremost we are proud that we were able to overcome the challenges we faced during the process of building the prototype, one of them being critical for our project to perform as we had desired at the beginning of the hackathon. Moreover, we were able to develop a fully functioning and interactive frontend for our prototype which can help the viewers to understand and test out what HOMA does.
We learned many different aspects of AI. We used AssemblyAI to add the speech to text aspect of it. We also spent time looking at OpenCV to get picture to text input as well; but that was not fully implemented. Finally, we researched aspects of text recognition. While we couldn’t find accessible tools, we used some of the ideas that text recognition is based on in an algorithm.
MacHacks 2 might be coming to an end but HOMA is just getting started. We have some aspiring future goals we as a team want to accomplish to make our app. We hope to increase the “intelligence” of HOMA such that it can solve even more mathematical business problems.