Progress report:

For the second week, I worked on refining what I had completed last week. In the case the clue that the child provides is not an exact match to the information provided in ConceptNet, but similar in meaning, I was able to get the system to return a guess on an object with a property that is similar in meaning to the clue provided. I was also able to narrow down the list of properties of each entity extracted from and image by cutting out repetitive terms. I also found code provided by a past intern in the form of an Android app and looked to that as a reference point on how I could narrow down the search space.

Some problems I encountered:

I was trying to figure out the most effective way to extract information from the user utterance and how I could utilize the NLP APIs. I was also concerned about potential issues with runtime and I had a difficult time understanding how NLP models generate QA pairs.