2.4 Discourse Ambiguity: When the meaning of any sentence depends upon the information given in the sentence just before it or right after it.
2.5 Anaphoric Ambiguity: When entities that were previously introduced are miss interpreted. Consider the example, the goat ran up the mountain. It was very steep. It soon got tired. The reference of “it” in the two situations is ambiguous. Steep is applied to a surface so “it” can be the mountain. [1].
2.6 Pragmatic Ambiguity: This ambiguity occurs when an entire context of a phrase has multiple interpretations. This is one of the hardest ambiguities for AI to understand. The user’s intention and sentiment have to be processed. This is a highly complex task for AI …show more content…
If the desired results are not achieved the algorithm must be changed and the results monitored again. Programmers were learning a lot about the problems of artificial intelligence through data analysis, but the machines themselves were only performing numerical optimization that optimized performance. The machines themselves learned very little. Deep learning is a new type of machine learning that only requires partial or no observation. This is achieved by changing the inputs to vectors. The concept of vectors are difficult for humans to understand but are easy for computers to understand. Words, phrases, and entire sentences can be represented entirely on vectors. Words and phrases that are similar can be found on vector planes that are close to each other. Words and phrases that are commonly used are repositioned on the vector to give them a higher weight. …show more content…
DISCRIMINATIVE DEEP ARCHITECTURES Discriminative deep architectures are deep networks for the purposes of supervised learning. Such architectures enhance the provision of discriminative power directly linked to pattern classification, and dependent on either Multi Level Perceptron (MLP) or neural network [3]. The Convolutional Neural Network (CNN) and Deep Stacking Network (DSN) are two of the most recent types of deep learning architecture [3]. Due to its inherent effectiveness, CNN has been widely applied to computer vision as well as image and speech recognition while DSN focuses on discrimination based on parallelization and scalable learning