I'm writing a script that converts my list of bookmarked words into SRS, with audio and image on it. Now, I'm writing a script to fetch an image from Google Image, but the images I got are mostly useless.
For example, when I searched for an image on a French word proviseur, I got the following list, which is a good representation of the word (principal).
But when I searched for a word nuire (meaning harm as a verb), the result is useless, with the top result being the conjugation table of nuire.
Or when I searched for the word fourni (well-equipped), the result is full of an island called fourni island.
Unfortunately, the most words are much closer to the latter than the former, once you leap over the very basic words (such as eat, dog, etc...) (but the results of nouns usually work well).
So how can I make an appropriate image for my SRS?