At this point in my project, I have achieved my first and most major goal of creating a product with a success rate of at least 75%. However, I am aware that there is potential for my product to achieve a higher success rate than it currently has. Out of the erroneous trials (the other 25%) many are missing the mark by merely a fraction of a percent. By this, I mean that my program is only more confident in an incorrect answer than it is in the correct name by an extremely small amount. This is likely due to more minor fine-tuning issues that could be alleviated relatively easily. As of right now, I already have some ideas as to where exactly in my code this fine-tuning may take place and I fully intend on exploiting the potential to raise the success rate. However, it is worth mentioning that the process of fine-tuning may take a fairly long amount of time. With each change that I would make to make an erroneous trial into a success, I would also have to run back through every other trial to ensure that ay previously successful trial has not been compromised. While it is time-consuming, this process shouldn't prove to be too difficult. Along with the fine-tuning adjustments I have also been looking into expanding my product's capabilities. Primarily, I have been researching various methods used to deal with background noise. Currently, my product would really only be successful in quiet environments, but this isn't always the type of environment a user may be in. So far, I have found that the two major ways of dealing with background noise rely on either physical methods, like simply using a directional microphone, or virtual methods which may involve two microphones and using software to identify and cancel out the background noise. Going forward I hope to continue to look into this possibility in tandem with making those minor adjustments.
Amol Kumar
Comments