Page 375 - AI Ver 3.0 Class 11
P. 375

consider some additional examples:


                      • You can bank on Ravi to finish the project on time. (bank here means rely on)
                      • I need to go to the bank to deposit money. (here bank means a financial institution)


                      • She likes to rock in her chair while reading. (move)

                      • The band rocked the concert last night! (perform)


                      • You can store files on a cloud. (online storage)
                      • A cloud covers the sky. (weather phenomenon)


                 It might only take you a moment to understand these different meanings, but an AI system could struggle with classifying
                 these elements correctly without a comprehensive understanding of language nuances and context.

                 Dealing with Classification Problems
                 Classification can be more challenging for an AI system than simply identifying tokens because so much of classification
                 depends on the context within the sentence.
                 Compare:


                      • I read a book about space. (outer space in universe)
                      • I need space to work. (personal space)


                 In both cases, the word “space” is used, but it has different meanings based on the context. An AI system must associate
                 the word with the correct context: outer space or personal space.
                 How does an AI system deal with this problem? Here are some ways:





                                         AI systems use machine learning techniques, such as supervised learning.
                         Dealing with Classification Problems




                                                 The AI system learns patterns and links between words, sentences, and their
                                                 meanings through a massive dataset of language usage and classification.





                                                 The  AI  system  improves  classification  accuracy  over  time  by  modifying
                                                 internal settings depending on observed patterns.




                                         To overcome the level of uncertainity and error, well-designed AI systems not only
                                         provide a response but also a confidence value, which indicates the system's degree
                                         of assurance in its classification.




                                                                   Leveraging Linguistics and Computer Science  373
   370   371   372   373   374   375   376   377   378   379   380