usa india home mail Request for Contact

Sentiment Analysis

Sentiment analysis is a special category of classification problems under natural language processing which lets you track the mood of your customers about your products or services. The objective of sentiment analysis is to extract the sentiment words from the unstructured text to classify them into positive, negative or neutral. A sentiment analysis system for text analysis includes understanding natural language and applying a machine learning algorithm to assess the sentiments of the sentences. It is done by calculating the weighted sentiment scores of words and special terms within the sentence or phrase resulting into negative, neutral or positive sentiment.

Sentiment analysis can be performed on different levels of granularity:

  • Document-level: The whole document is classified into either positive or negative according to the sentiment expressed in the text.

  • Aspect-level: The overall opinion is considered to know the sentiment instead of looking at language constructs in paragraphs,phrases or sentences.

  • Phrase-level: Sometimes people prefer phrasal words to be more expressive to show their liking or disliking. Phrase-level sentiment analysis handles phrasal words carefully to judge the sentiment

  • Sentence-level: Every sentence is classified as positive, negative or neutral based on the words contained in the sentence.


Some of the challenges faced by Sentiment analysis are:

  • Context and Polarity: Sometimes the words which express the sentiments in the sentences are said in continuation of the previous sentences. The contextual understanding of the words by looking at preceding sentences plays a vital role in describing the polarity of the words.
    E.g., I am using the product for five years. It has started giving some problems now. (Positive)
    The product did not seem right initially. But now I have understood howto use it. (Positive)
    I thought I don’t know how to use the product initially. But it’s like that only. (Negative)

  • Indirect Negations: The negation words such as no, not,etc., can be used to show negative sentiment in the text.
    E.g., I like to watch this movie. It is not boring.
    Here, Boring is a strongly negative word. But itbecomes positive due to preceding not.

  • Comparisons: The polarity of the sentence is sometimes strongly expressed in terms of traditional product comparisons.
    E.g., The Battery life of Nokia phone is much better than a Samsung phone. This represents a positive review for Nokia and negative for Samsung.

  • Correlation Detection: Many sentences indicate strong sentiments. But finding the source of the sentiment is difficult. Detecting correlation to specific keywords makes it possible.
    E.g., When I read the book “I too had a love story,” I really want my husband like him.

  • Co-reference Resolution: In many sentences, the sentiments are expressed strongly, but it refers to subjects mentioned in previous sentences. Identifying what a pronoun or a noun phrase refers to gives you better control over analyzing the sentiment.
    E.g., We found the product online, ordered it and went to dinner; it was awful.What does "It" refer to? Co-reference resolution may be useful for the topic/aspect-based sentiment analysis. Co-reference resolution may improve the accuracy of opinion mining.

  • Domain Dependence: Same word may represent different sentiments in different domains. Association of words to specific domains make it positive in one domain and negative in other domain.
    E.g. The unpredictable performance of the product catches my attention- Positive. The performance of the product is unpredictable -Negative.

  • Sarcasm Detection: Sarcasm expresses negative sentiments using positive words. It implicitly conveys alot about the mood of the person.
    E.g., Nice perfume. You must marinate in it.
    Although it’s a positive sentence, it is indirect negative feedback about the perfume.

  • Subjectivity and Tone: The sentiments are sometimes expressed for specific subjects, and the tone of the words makes it positive or negative

  • Thwarted Expressions: Sometimes, only a few words of the sentence present the overall sentiment.
    E.g., Good color, light-weight and handy, but the hair dryer is not good.
    Even though the first three words in the sentence are positive, the overall review is negative due tothe last words.

  • Emojis: This has become a common practice to use emoticons in the text. Understanding the emotions attached to every emoji may add a lot of weight on the text.
    E.g. The product for me is (negative)
    The product does a good job (positive)


The above challenges can be dealt with by precisely considering the sentiment of the individual sentence in a group of sentences.

  • Every word in each sentence is measured for its lexicon value which isdetermined by two factors viz. polarity and intensity.

  • A compound score is calculated for the set of sentences by evaluating the polarity and intensity of all the words.

  • This compound score can vary in the range –n to +n representing –n being the most negative sentiment while +n being the most positive one with 0 being neutral sentiment.

  • The more the intensity of words, the higher the compound score.

Use Cases

Sentiment analysis is also referred to as emotion AI, which involves analyzing views of people from the written text in order to understand and gauge reactions. AI and sentiment enabled applications lets you effectively measure the sentiments ofyour customers for their mentions. Here are some use cases as in why businesses need sentiment analysis.

  • Case Management

    Once a product is sold to your customers, they can sometimes come back to you in case if they find any problem in your product(s)/service(s). This is usually done by raising a case ticket or by sending a complaint email to the company.These cases and emails can be taken care of by the company based on different criterions. Sometimes it is done based on the level of urgency looking at the subject line of the email or simply in chronological order of receive date.But if this is done based on the severity of the complaint, the customer's experience can be improved. This can be achieved by sentiment analysis of text written both in the subject and the body of the email. More severe cases can then be looked into first than others. The received emails can be analyzed for their sentiments and accordingly classified on a scale of 5 as follows:

  • Sentiment Analysis as Lead Generation Tool

    Lead generation is one of the most important concerns of any business. It is all about bringing more customers to your sales funnel. As the latest trend, companies have started a social media platform to generate leads for them. Sites like Facebook, Twitter, LinkedIn, and Instagram are preferred due to cost-effectiveness. Every mention on social media becomes essential as it can act as a way of identifying potential clients for your business.
    A careful analysis of these mentions may give you an idea about:
    o The interests of people.
    o The degree of liking of your product(s).
    This information enables you to know how interested people can be converted into your customers.

  • Sentiment Analysis as Product Performance Analyzer

    Existing customers of your product find it easy to show their liking or disliking about your product. Having an online presence,the customers give their feedback about your product on social media platforms. This feedback allows you to know the following:
    o Identifying the pain points.
    o Identifying expectations of customers.
    It also gives you an understanding of the features to be added/improved in your product based on customers’ expectations. At the same time, these feed backs become a driving factor for lead generation.


A robustSentiment Analyzer lets you perform sentiment analysis of your product givingyou a better understanding of the following:

  1. Voice of the Market: This determines customers’ feelings towards products or services of competitors.Timely information about the voice of the Market gives you a competitive advantage. Such information at an early stage of launching the product to know target customers. This real-time information also helps you design new marketing strategies in case of a new product and improving product features in case of existing product.

  2. Voice of the Customer: This determines the individual customer’s opinion about your products or services. It does so by analyzing the reviews and feedback of individual customers. This lets you manage Customer Experiences over time. Extracting customer opinions helps in identifying both functional requirements of the product like its features and non-functional requirements like performance and cost.

  3. Online Presence: This determines your positive or negative online presence based on users’ experience about the product quality. The customers assess your product by assigning ratings and feedbacks. Prospective customers can view opinions and recommendations about the features of your product and their experience using your product. Sentiment analysis helps in determining how a company’s brand, products or services are perceived by community online.

  4. Brand Reputation Management: This can work to manage your brand reputation in the market. Opinions of customers or competitive companies can damage or enhance the reputation of your company. A graphical summary of the overall product and its features allows you to take strategic decisions in the long and short term.

Functions and Features

The Sentiment Analysis model deals with some of the typical cases including proper handling of sentences with:

  • Contraction as negation

  • Punctuation to signal increased sentiment intensity

  • Capitalization for more emphasis

  • Acronyms

  • Emoticons

  • Consideration of preceding and succeeding terms