As you will see I am totally ignorant of Computational linguistics and NLP. However, I am a novice linguist who has developed a fascination in how you can use NLP for linguistics.
This is my question; I was reading a book the other day, and as I was reading a sentence I mispronounced "row" (as in to row a boat) as "row" (to argue). It was only clear to me when I saw the connecting part of the phrasal verb "out" as in "he wanted to row-out to the island" what the meaning was. This was in part because the - (dash) was ommitted. Also, because I was tired.
I then asked myself this question; "How does a program determine if "row" means to argue or to row a boat? Does it read in a linear fashion, word by word or does it read sentence by sentence? Can it predetermine according to the previous sentence or paragraph which meaning of "row" it is likely to be? For example, if the main verb in the previous sentence is "climb", is the program told to understand that the following verb is likely to be a verb that is also classified to express action or physical doing? I imagine it depends on how the program is programmed? Is there a common or dominant formula for how a program would understand or tackle this potential semantic problem?
Any advice on how I can know more about these things would be very interesting to me.
