Social Media Sentiment Analysis is a field of study with a vast number of applications. One important application is analyzing customer purchasing behaviors using the results of social media sentiment analysis which is a great tool that decision-makers can utilize. There are several studies conducted in this field. This paper presents the results of a systematic literature review conducted on the existing studies which would be beneficial for developers and researchers interested in this field. This is a preliminary SLR since only articles from repositories of ResearchGate, Emerald Insight, ScienceDirect, and IEEE Explorer are studied for this study. Most relevant studies derived through specific inclusion and exclusion were investigated to analyze approaches and methods used, results, limitations, gaps, and future recommendations. The results of this study suggest that from machine learning and lexicon-based approaches, machine learning produces more accurate results in sentiment analysis. Specially Naive-Bayes and Support Vector Machine methods. Hybrid approaches are the most accurate. Only a limited number of researches have been conducted utilizing both words and emoticons in the text. Studies on social media sentiment analysis on customer purchasing behavior are conducted to produce insights into the impact of cultural differences, the period, brand loyalty, and various other aspects on customer sentiments.
1. Introduction.
Sentiment Analysis
Sentiment analysis is also known as opinion mining and is a subdomain of Natural Language Processing. It aims to derive the writers’ opinions, evaluations, emotions, or attitudes from a text including the degree of positivity or negativity of the sentiment. (1,2,3). According to [4] it also attempts to identify is a text is subjective or objective. Sentiment analysis provides is applied for a variety of purposes. Movie Reviews [5], product reviews [6],[7], blogs, and news ([4], [8]).
Save your time!
We can take care of your essay
- Proper editing and formatting
- Free revision, title page, and bibliography
- Flexible prices and money-back guarantee
As stated in [8] sentiment analysis can be conducted using the following approaches
- a) Machine learning based.
- b) Lexicon based.
- c) Keyword-based.
- d) Concept-based.
The following issues exist in sentiment analysis according to [9]
- a) The meaning of words or sentences varies according to the domain it is being used.
- b) Sentences, especially if they are interrogative or conditional, might not produce a sentiment but a single word might produce a sentiment.
- c) Sentiment of the sentences without sentiment words is hard to identify.
- d) Sometimes a single text might generate the opposite sentiment as a result of sarcasm.
- e) The sentiment of a sentence depends on the person. (a single sentence can have different sentiments according to each person).
- f) Language issues depend on the place it’s been used.
Social Media Sentiment Analysis
At present social media platforms are spread everywhere within the globe. The digital age that we are currently living in has nourished the fast growth of social media [10]. Approximately 2.82 billion people were actively using social media in 2018 worldwide. This number will be increased to 2.93 billion in 2020. Facebook, YouTube, WhatsApp, Facebook Messenger, WeChat, Instagram, QQ, QZone, Douyin/Tik Tok, Sina Weibo, Reddit, and Twitter are a few popular social media networks. Users freely express their ideas, attitudes, emotions, and opinions about anything and everything in these social media platforms[29]. This creates a massive amount of user-generated content that falls under the category of big data. When the sentiments contained in this huge amount of text data are analyzed properly the results produce valuable insights that can be applied in several fields.
As previously mentioned, the vast number of applications of social media data analysis results is facilitated by the fast expansion of social media around the globe[11]. Product recommendations [12], stock price prediction [13] [14], sales prediction [15][16][17][21], election results prediction[18], public health planning[19], economic forecasting[20], political analysis [22][23], disaster predictions[24].
However, social media sentiment analysis includes its’ own benefits and limitations. For example, gathering data is relatively easier and low-cost. There is content freely available about any topic. On the negative side, in some geographical areas, especially in developing countries social media does not represent the opinion of the general public as the majority of the population using social media tends to be younger and socio-economically privileged [25]. In addition, language changes are harder to analyze. Social media data is also unstructured, high volume, and noisy [26] thus traditional methods are powerless to analyze them.(Abraham,[27])
Social Media Sentiment Analysis for Customer Purchasing Behavior
Social media sentiment analysis provides valuable insights related to customer purchasing behaviors. Social media is so integrated with human lives, serves as the main method of communication, and affects all kinds of human activities.[28] As mentioned in [29] most users express their opinions about brands and services on social media platforms. Users freely and frequently create and share content on what they purchase, what features they like or dislike, how disappointed or loyal they are to a specific brand etc.
When analyzed properly, this user-generated content plus content created by other brands which are in the form of comments, posts, tweets, or retweets uncovers hidden knowledge that decision-makers can utilize for the success of their business.
Emotions are what drive humans in everything they do, including making purchasing decisions. The emotion, opinion, or sentiment they have about a specific product or service determines whether they make a purchase or not. Their opinions are affected by both leading figures and the general public thus their decision-making process is also affected by both parties[30]. Thus closely monitoring social media data enables firms to be informed of shifts in customer sentiment and react according to reduce harmful effects. They can also utilize it as a performance evaluator: to identify what products and services need to be improved, the degree to which customers' needs are met, and the successfulness of their market strategies. Social media sentiment analysis is also utilized as a market research tool by analyzing customers' sentiments about competitors. [ He, W., Wu, H., Yan, G., Akula, V., & Shen, J. (2015). A novel social media competitive analytics framework with sentiment benchmarks. Information & Management, 52(7), 801–812.] analyzed social media comments for market research that identified leading companies in the technology sector. Social media platforms provide companies with knowledge of the best strategies to be applied [31]. Hidden patterns and knowledge from social media sentiment analysis results will provide firms with a competitive advantage[He, W., Zha, S., & Li, L. (2013). Social media competitive analysis and text mining: A case study in the pizza industry. International Journal of Information Management, 33, 464–472. doi:10.1016/j. ijinfomgt.2013.01.001] Need to add more papers exactly about SMSA for CPB
2. Research Method
This study follows the methodology given in [32] to conduct the SLR. The following are the steps followed.
2.1. Strategy of Literature Search
As mentioned earlier, plenty of studies have already been conducted in this field. The application of a smart searching strategy allows us to retrieve as many of them as possible.
Social Media Sentiment Analysis “social media sentiment analysis”
Review “systematic literature review”, “systematic review”, “systematic mapping”, “mapping study”, “systematic literature mapping”
Search string (“Social media sentiment analysis”) AND (“systematic literature review” OR “systematic review” OR “systematic mapping” OR “mapping study” OR’ ‘systematic literature mapping”)
Customer Purchasing Behavior “Customer purchasing behavior”
Review “systematic literature review”, “systematic review”, “systematic mapping”, “mapping study”, “systematic literature mapping”
Search string (“customer purchasing behavior”) AND (“systematic literature review” OR “systematic review” OR “systematic mapping” OR “mapping study” OR’ ‘systematic literature mapping”)
2.2. Inclusion and Exclusion Criteria
Studies to be included and excluded from this review were determined under the following criteria.
-
- Inclusion Criteria
- IC1- Publications from repositories: ResearchGate, Emerald Insight, ScienceDirect, and IEEE Explore were fetched.
- IC2- Publications published from XXXX to XXXX were included
- IC3- Studies related to only social media sentiment analysis or identifying customer purchasing behaviors and studies related to both social media sentiment analysis and customer purchasing behavior were included.
- Exclusion Criteria
- Studies without an abstract, written in a language other than English, and summary or editorial studies were excluded.
- EC2-If a study has multiple versions, all except the latest were removed.
- EC3-Any duplications were removed.
2.3. Research Questions
In this review, the following research questions are addressed.
- RQ1: What are the aspects that have been covered in the existing research regarding social media sentiment analysis for customer purchasing behavior?
Explores the aspects of SMSA for customer purchasing behavior that have been already covered to a satisfiable level and what aspects are less investigated
- RQ2: What are the issues/limitations the previous researchers had to face in this field of study?
To get a fair idea of the main problems and barriers the former researchers had to overcome. Another goal is to identify what are the limitations and where future research should focus on.
- RQ3: What are the applications of knowledge generated in previous research?
This question investigates the reasons for conducting SMSA for customer purchasing behavior.
- RQ4: What are the technologies/methods/approaches used to conduct this type of research?
Discovers the technologies/methods/approaches currently used to conduct this type of research. This question can be used to identify strengths and weaknesses in current technology and guide toward novel technology to overcome any gaps.
- RQ5: What are the main conclusions drawn from existing studies?
Highlights the basic knowledge other researchers have gained through their research. This knowledge is essential for any new research to be conducted in this field
- RQ6: What are the types of studies a) present an entirely new approach or a methodology b) utilize an existing methodology c) extend an existing methodology?
This question is similar to RQ4. It aims to classy existing studies based on approaches and techniques used so future research can focus more on studies that introduce novelty.
2.4 Data exaction and synthesis method
The following steps were followed to exact and synthesize data from research papers.
- Articles were fetched as in IC1 using the search strings mentioned in section 2.1.
- Articles complying with IC2 and IC3 were shortlisted
- EC1, EC2, and EC3 were applied to shortlist the more suitable articles list.
- Each selected article was analyzed under the research questions mentioned in section 2.3. with the aid of a form.
- All the results are arranged and sorted.
4. Results and findings
- RQ1: What are the aspects that have been covered in the existing research regarding social media sentiment analysis for customer purchasing behavior?
Former studies are conducted on identifying the best method between different methods that can be used[https://www.researchgate.net/publication/331111847][https://www.researchgate.net/publication/337151689], creating more efficient and accurate frameworks and approaches, different tools that can be used [https://sci-hub.tw/downloads/2019-07-18/50/10.1007@978-3-030-14070-068.pdf#view=FitH], different applications of analysis results. A very limited number of researches have been conducted on predicting purchasing behaviors
- RQ2: What are the issues/limitations the previous researchers had to face in this field of study?
If research is conducted using a lexicon-based approach then the final results heavily depend on
the words of the lexicon. A language is a complex intellectual that is constantly changing. The meanings of words depend on the context being used. Researchers in this field of study must address all these issues in natural language processing. On the other hand in some countries social media is not used by the entire population, mostly younger and socioeconomically benefited people tend to be active social media users[25]. So making conclusions about the entire customer base is questionable.
- RQ3: What are the applications of knowledge generated from previous research?
Results of analyzing social media sentiment for customer purchasing behavior are applied in every place where awareness of the sentiment or the opinion of the customers is critical. Firms tend to pay more attention to gathering data about their customers’ opinions because of this reason[Liu, B. (2012). Sentiment analysis and opinion mining. Synthesis Lectures on Human
Language Technologies, 5(1), 1–167.] For example [Zhong 2019] analyses differences in customer sentiment in different cultures for the same product- this result is useful to marketers and decision-makers. The study [Ibrahim] analyses customer sentiments promptly to discover the effect of events that occur on the customer's opinion. Exploring trends in customer sentiment is also very useful[Ibrahim 2019]. Measuring consumer confidence via customer sentiment is another application[shaaya2019]. Sales performance can be analyzed using customer sentiment [Kim 2019]. Another very useful application is predicting customer purchasing behaviors using customer sentiment analysis results.
- RQ4: What are the technologies/methods/approaches used to conduct this type of research?
Various researchers have followed different approaches when conducting this type of research utilizing machine learning-based methods (mostly Naïve-Bayes, Support Vector Machine, K-mean, K-nearest neighbor), lexicon-based methods, or a combination. Some researchers attempt to increase accuracy by removing fake content[kaffma-2019], some aim to identify and remove sarcastic content to improve accuracy, and some simply apply multiple methods to prove which method generates the best results[putra, paper icon eeimiz-2018,1512,sent-2019], some conduct social media sentiment analysis time-based manner[Zhong. 2018], some even attempt to discover reasons behind the sentiments, etc. Need to add more approaches. Programming languages R [Liske, D. (2018). Tidy sentiment analysis in R. https://www.datacamp.com/community/tutorials/sentiment-analysis-R.]and Python are adopted by many researchers as they have inbuilt packages
- RQ5: What are the main conclusions drawn from existing studies?
Even though this field of study has been discussed for a long period, the highest accuracy of the sentiments generated varies between 80% to 86%. Despite the number of research being conducted in this field, still there is room for novelty and aspects that need improvement. Most studies only utilize words for generating sentiment, and only a limited number of studies attempt to utilize both words and emoticons that appear in almost all user-generated content on social media. Regarding the type of approaches being used, machine learning approaches are more accurate than lexicon-based approaches and hybrid approaches produce the highest accuracy.
- RQ6: What are the studies that, a. present an entirely new approach or a methodology b. utilize an existing methodology c. extend an existing methodology?
Researchers have been studying social media sentiment analysis for a long period, and multiple tools and technologies have already been invented. Hence most studies tend to use existing technologies or extend them to produce results. However, new studies present new approaches to overcome the barriers and limitations related to accuracy that still exist in this research field.
Conclusion
Social media sentiment analysis for customer purchasing behavior is a very broadly discussed topic by researchers, studies have been conducted continuously utilizing various methods and techniques. In this SLR XX, several studies from libraries: ResearchGate, Emerald Insight, ScienceDirect, and IEEE Explor were initially taken into consideration and XX was shortlisted for complete analysis under the five research questions mentioned in section 2.1. The following are the main conclusions driven,
- A limited number of researches are conducted on using both words, emojis, and emoticons to increase the accuracy of sentiment analysis, detecting sarcasm and removing fake content to increase accuracy, identifying the reasons behind sentiments, predicting sentiments, how to fix issues that arise in customer purchasing behaviors as an impact of sentiment shifts, using sentiment analysis identify characteristics of consumer groups
- Most of the researches only focus on classifying sentiment as negative, positive, or neutral. A smaller number of researchers declare the intensity of negativity or positivity.
- The issues in natural language processing techniques are all in sentiment analysis.
- Studies that used hybrid models sentiment analysis and extended lexicons achieved the highest accuracy.
Stuck on your essay?
