Refining my Product Proposal with new ideas!

Last week and this week, apart from finding my mentor for ISM this semester, I have been really trying to get a better understanding on how I can create a final product using machine learning and solve real world problems that are close to me. In fact, I found a video posted by Google's TensorFlow platform that talked about how we could spot problems in our daily lives and use machine learning to solve them. The three main factors that we took into consideration were making multimedia searchable, understanding language, and converting between data types.

Basically, by making multimedia searchable, we could navigate and find specific instances of our lives very easily instead of manually going through each of the videos to find that moment. This example was very simple and pretty unique as it solved a real world problem and was very intriguing for me. Another aspect that we must consider is semantic search to understand language or search queries that essentially have the same meaning but are written in different words. Machine learning has the solution to this by using word embeddings, assigning words in a given space to provide results to similar word queries. This turns out to be very useful as we could obtain results to specific queries even when they don't exactly match the keywords of our model. Another example of solving machine learning problems is converting from PDFs to audiobooks by using machine learning. To do this, we would need to identify sections, extract text, and ultimately convert that text to speech using several different APIs. You could also translate text from videos and audios to help users! I think that all of these examples were very useful for me to learn and I am starting to get a better idea of what I can do using pre-written APIs and my own knowledge to solve problems.

So see you next week, same place, same time.