Sentiment Analysis

Home  >  Blog

Sentiment Analysis

What is Sentiment Analysis?

According to Oxford Dictionary, the definition of Sentiment Analysis is the process of computationally identifying and categorizing opinions expressed in a piece of text, in order to determine whether the writer's attitude towards a particular topic, product is positive, negative, or neutral.

Sentiment Analysis (Feeling/Sentiment analysis) usage areas

Use of sentiment analysis is often implemented in social media in order to help marketing and customer service teams to define consumers' feelings (categorizing text as positive, negative or neutral). It can also be used in product analysis to discover whether consumers are satisfied with a product or not. Similarly, a company can implement sentiment analysis in order to calculate the effect of a new product, advertising campaign or consumer's response to latest company news in social media. Companies can measure if after sale customer service support meets customers' needs or not. Politicians can carry out measurements on the responses of the citizens to the campaigns during election periods.

Difficulties in Sentiment Analysis:

People express their ideas in complicated ways so their views become hard to understand. Sarcasm, irony and implication can misdirect sentiment analysis.

For example, although the sentences in two pictures are the same, they are extremely different in terms of meaning. Thus, Sentiment analysis can be wrong.

Therefore, analyzing basic and focused ideas such as product, book, movie and music is easier. Methods to use for sentimental analysis;

  1. Manual processing

It is a method in which people manually evaluate and score the sentences they read. It is the method with the highest preciseness. It can still not create 100% true results. It has been more difficult to use the method in the recent years when social media has a rapid growth.

  1. Keyword processing

Keyword processing algorithms allocate a positive or negative value to a word; afterwards it scores and gives a percentage score to the text. For example; positive words, great, like, love or negative words terrible, dislike.

Figure 2: Keyword processing (Positive/Negative word groups)

The advantages of the method are that it is fast, predictable and cheap to implement and operate. Additionally, in the event that some words can have both positive and negative meanings, the word will be included in both positive and negative groups causing a miscalculation in the evaluation. Different people can also assign positive or negative different percentages to a word.

  1. Natural Language Processing (NLP: text analytics, data mining, machine learning linguistics) The algorithms which are generally prepared in Python language can assign positive/negative/neutral values to the sentences by using their grammar skills.

For example; Positive Sentence. Input:

{ "document": "I really like eating ice cream in the morning!"} Output:

{ "sentiment": 0.474, "document": "I really like eating ice cream in the morning!" } Negative Sentence


{ "document": "I really hate you, you are the worst!"} Output:

{"sentiment": -0.855,"document": "I really hate you, you are the worst!" }

As a result, carrying out Sentiment analysis is an extremely hard process. In the recent years, algorithms with machine learning skills which use Natural Language Processing technology have been developed so accuracy rate in Sentiment analysis increases.

Murat Kerem Çelenli

TAV Technologies BI/MIS and Oracle Systems Coordinator