![]() |
|
Philosophy Religions, schools of thought, matters of importance and navel-gazing |
|
Thread Tools | Display Modes |
![]() |
#10 | |
Draco dormiens nunquam titillandus
Join Date: Nov 2005
Location: Melbourne, Vic
Posts: 316
|
Quote:
![]() Here are ten words, each put into grammatically-complete sentences (so an adequate amount of syntactic context is present). Each word is either a noun or a verb with two meanings and each meaning has a different pronunciation. The sentences do not have any semantic context supplied. Each sentence has a short question associated with it. If the sentences were spoken to you, you would be able to answer all the questions correctly. However, the sentences are written, not spoken. This makes the meaning ambiguous; you cannot answer the questions because as written both answers are plausible. He is putting on the first.
To gain some insight into how the ambiguity can cause difficulty, it is instructive to experiment with text-to-speech engines. Text-to-speech engines can use syntactic context to disambiguate, but they cannot make use of semantic context because it is very difficult - if not impossible - to program computers to understand semantic context with 100% accuracy, and certainly not possible with the current state-of-the-art in desktop operating systems. If you have Windows XP or Windows Vista, you can access the built-in text-to-speech engine in this way: Control Panel, then Speech. There is a prompt there that says: "Use the following text to preview the voice." If you paste the sentences into this prompt, and then click the button that says: "Preview Voice", it will read it out. However, the sentences I provided do demonstrate the limitations of the technology. For example, the first sentence I gave reads as follows: "He is putting on the first." The text-to-speech engine assumes that the verb is "put", not "putt". Even if you add the word "green" to the end of the sentence (which provides some semantic context for golf that you can disambiguate as a human), the text-to-speech engine still says it as if the verb was "put". This shows that computers (or, to be more precise, Microsoft's text-to-speech engine) cannot understand semantic context very well. Syntactic context is different. Computers understand this relatively easily. If you have it read the text: "We estimate to make an estimate." (a little contrived but it demonstrates the point adequately), the text-to-speech engine reads both occurrences of the word "estimate" correctly even though the two instances are pronounced differently (the last syllable of the verb has a clearly-pronounced vowel and the last syllable of the noun has a reduced vowel).
__________________
Ur is a city in Mesopotamia. |
|
![]() |
![]() |
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
Thread Tools | |
Display Modes | |
|
|