re: Introduction to Natural Language Processing, Part 1. VIEW POST

FULL DISCUSSION
 

Thanks for the crash course into Parsey McParseface. Out of curiousity, can Parsey McParseface parse "Parsey McParseface"?

Also do you have any specific reasoning behind the assertion that "Computers can’t truly understand the human language". Is there a benchmark that you think would be difficult, or is this an epistemologic hurdle?

 

Hi Andrew, Interesting question. Parsey McParseface is trained with a lot of English words. If the word 'Parsey McParseface' is already labeled into it, it can parse it. But, if it's a new word, it will try guessing to its nearest meaning, but can be embarrassingly wrong few times. It has 94% accuracy, so I guess Parsey should parse it.

"Computers can’t truly understand the human language", we can train a model and it can do tasks on the basis of the data it has been fed. But it cannot understand and make a sense of sentences like we as humans can do. Human language has a lot of ambiguity, it's difficult to get the 'correct' meaning of a sentence. We as humans can figure out the variations but computers can't. Yes, making computers understand this will be difficult, but parsers like Parsey might try leveling down such ambiguities.

code of conduct - report abuse