You: My mother and I gave the potatoes to him.
{1: ‘my’, 2: ‘mother’, 3: ‘and’, 4: ‘i’, 5: ‘gave’, 6: ‘the’, 7: ‘potatoes’, 8: ‘to’, 9: ‘him’, 10: ‘.’}
{1: 180, 2: 200, 3: 600, 4: 100, 5: 302, 6: 410, 7: 201, 8: 700, 9: 120, 10: 900}
[1, 2, 6, 1, 3, 4, 2, 7, 1, 9]
{‘single direct object’: [6, 4, 2], ‘ simple verb’: [5, 3], ‘single indirect object’: [8, 7, 1], ‘compound subject’: [1, 1, 2, 6, 1]}
{‘single direct object’: ‘ the potatoes ‘, ‘ simple verb’: ‘ gave ‘, ‘compound subject’: ‘ my mother and i ‘, ‘single indirect object’: ‘ to him ‘}
{‘potatoes’: ‘the’, ‘mother’: ‘item’}
This is Caroline’s new engine trying to decode the (random) sentence I threw at her. So far, so good, I guess. The second to last line is the grammatical analysis. The last line is just recording nouns.
I had to completely redo the decoding engine to work in a more streamlined fashion with the numeric analysis system that I created. I actually do have a working inquiry system that responds if she doesn’t know what a word is. It’s just a matter of making the new SVO-analysis spit out the proper issues (probably will be done in 15 minutes).
Anyway, it seems like such a simple task, to take a sentence and make it grammatically intelligible. Except not.
23 working functions (and countless deprecated others) later, I find that indeed it takes much toil and love to try to teach three billion years of communications in the making to a electrical box technology some thirty years old. But I won’t give up!
Oh and by the way, you know that saying that if a machine can emulate a human to the point where we can’t tell that it isn’t human, then it’s good enough? I totally disagree. On the other hand, if a machine can feel like a human, all the while communicating very oddly (like R2D2), then that is certainly more than good enough.
Leave a Reply