Page 374 - AI Ver 3.0 Class 11
P. 374

Emotion Detection and Sentiment Analysis

              Sentiment analysis and emotion detection are two NLP techniques
              that use human language to categorise people's thoughts, attitudes,
              and feelings. Sentiment analysis is a text classification technique that
              determines if subjective information is favourable, negative, or neutral.
              Emotion detection employs machine learning to examine complicated
              emotions such as fear, anger, sadness, love, and frustration.
              At first look, sentiment analysis and emotion detection may appear
              to be the same concept particularly for individuals without a scientific background. However, they are not synonyms, so
              what is the difference between sentiment analysis and emotion detection?

                  Factors                 Emotion Detection                            Sentiment Analysis

                 Definition   Identifies various human emotion types.     Measures the intensity of an emotion.
                              Seeks  to  identify  the  emotions  expressed  in  Sentiment  analysis  seeks  to  categorise  data  as
                 Examples
                              texts, such as happiness, rage, and grief.  positive, negative, or neutral.
                                                                          Reading  social  media  content,  customer  service
                Applications  Assessing user ratings, survey comments.
                                                                          chats etc.
                                                                          Uses a sliding scale between positive and negative
                AI Training   Can be taught to classify emotions.         e.g. strongly disagree, disagree, neutral, agree, or
                                                                          strongly agree.
                              Identifying  emotional  tokens  to  understand
                  Purpose                                                 Evaluating the overall tone or sentiment of text.
                              context.

              Sentiment  analysis  focuses  on  the  overall  sentiment  expressed,  whereas  emotion  AI  seeks  to  distinguish  and
              categorise distinct emotions. Thus, they work together to provide insights into the subjective components of human
              communication.


                      Classification Problem

              In Natural Language Processing (NLP), classification problems include assigning a label or category to a piece of text,
              sentence, or phrase based on its content. For example, Emails can be classified as spam or non-spam, tweets as positive
              or negative, and articles as relevant or irrelevant to a specific topic.
              Human language often includes terms that are vague or have multiple meanings. This leads to a classification problem.
              Understanding these terms can be challenging, especially for AI systems that rely on context to accurately classify given
              elements within sentences. Let us understand this with the help of an example. Consider the following statements:

                   • He’s a really cool person. (Statement 1)

                   • The weather today is cool. (Statement 2)


              In the above statements the phrases “cool person” and “cool weather” show the ambiguity of language, which poses a
              classification problem. In statement 1, “cool person” can refer to someone who is admired for their style or demeanour.
              Whereas in statement 2, a “cool temperature” refers to a lower, bearable level of heat. The challenge arises because the
              word “cool” has been used in two different ways.
              Just as with humans, it can be tricky for an AI system to correctly interpret these meanings without proper context. Let’s



                    372     Touchpad Artificial Intelligence (Ver. 3.0)-XI
   369   370   371   372   373   374   375   376   377   378   379