The Role of Social Media in Today’s World

Discuss about the Sentiment Analysis in Carrabean Social Media for Caribbean People.

In today’s world the importance of social media has increased manifold. It is considered to be the information base which is relevant to all the users across the globe. It is used by a community of individuals, students, colleges, schools, universities and even business companies for the promotions of their products and services. It serves as an useful tool to analyse the thinking and preferences patterns of its customers. Thus, the main aim of sentiment analysis is to ascertain the attitudes and thinking of the people who regularly use the various facets of social media in their daily lives. This reports projects a semantic sentimental approach in order to identify the various preference which are projected by social media in accordance with the choices and wishes of its users. The different data which have been derived from the various sources, which have been obtained about the Caribbean people have been presented.

Social media, today gives the opportunity to its various users to publicise their personal as well as their professional information with the rest of the world at any time of the day. Instagram helps in posting of private as well as other photos of the users, whereas Facebook helps people to make their photos, comments and like preferences to go viral. This serves as an important tool for today’s companies to exploit. Due to the sharing of information over social media becomes easily accessible, the companies can easily predict the future patters on the basis of the past data and can make a product or come up with a service which exactly caters to the demands of the public in general. Proper understanding and follow up of the choices, preferences, comments, likes and other useful information of the general public is very important for both the public as well as business companies and policy creators. (Newman, 2017.). For example Google Company silently takes into account and tracks the various political issues around the world and the dangerous diseases plaguing the world today, in order to help the policy legislators and creators to eradicate these diseases and abort the influx of any terrorist mishaps or other political tensions.  (Gillespie, 2014). Sentiment Analysis, thus can be defined as the task of ascertaining or deriving the sentiments of the various texts and speeches of the persons in general.  Sentiment analysis is actually much more profound and deep than the commonplace of habit of most people who regularly use hashtags, to emphasise them.. In this report, a semantic analysis and approach has been employed by creating an ontology of sentiment, containing a group of various words which express separate groups of posts of Facebook which are called Facebook posts or statuses. These are taken as the examples of social media.

The Use of Social Media by Business Companies

There exists no limit to the amount of work, which exists today and helps in using a variety of techniques and procedures for the purpose of sentiment analysis. One of the works of Mr Pang and his colleagues on sentimental analysis, is actually one of the most earliest and iconic approaches, which had applied sentimental analysis upon online review of movie using the concept of machine learning.  Various associated and closely linked works like that of Lai in 2010, had applied sentimental analysis in Facebook by applying  various machine learning procedures in order to segregate the sentiments associated with those posts.. The research by Conover in 2011, had showed that social media giant Facebook could also be used as a platform for political contemplation and cognition. Additionally, the work also makes use of sentiment glossary and the results prove that the derivation could can create similar results like traditional polls of election.  This implies that if they expand the results of the sentiment glossary and lexicons, the results would be more effective as well as efficient. There have not been many studies which have been performed on the Caribbean social media. Only a few have been performed. For instance, in accordance with the study of Abdul Mageed in the year 2014, had solely focused on the product and movie reviews. As a consequence, the work of Mr Abbasi back in 20008, make use of algorithm of genetic nature for the detection of sentiment in both Web forums as well as in English. These are also based on document level. They make use of both the stylistic as well as the syntactic characteristics and features but they refrain from using any morphological characteristics and features. Even the task performed by Shokri back in 2012, profoundly states, the varying impact of preliminary processing on sentiment analysis of the various posts of the Egyptian dialects and languages.

In this part, the various details of the miniature model for the purpose of analysis has been   presented. The figure mentioned below portrays the different parts of the model. The three initial steps portrays the posts for classification.  Most of the contributions have been presented in the below mentioned steps, the ‘Classify’ step which primarily depends on the sentiment ontology. Thus, it can be said that we show and portray the reference of this in the miniature model in the subsequent paragraphs. In the paras to be followed, at first we initially introduce the work related to the sentiment ontology and after it, the project has proceeded to understand and explain the project’s main work conducted upon sentiment ontology and the explanation the ways to be used in order to analyse the posts.

The Importance of Understanding Public Preferences

The first and foremost step of the miniature model is creating and producing a collection of posts which are related to a particular topic or the brand name. Various approaches and ways tends to exists for the various social media websites and platforms. An example can be taken in this case. For example, Facebook Archivist could be used to download the various kinds of Facebook posts using several approaches. It can be done by using a Boolean search, a hashtag or a Facebook user or a query of complex nature or just a term. Similar kinds of instruments exists for various other platforms.

An important problem which is being consistently faced while dealing with the various facets of social media is the writing pattern of the posts. Mostly majority of them have an informal style attached to them. These posts are mostly written in a very informal and colloquial style which reflects a local dialect. They are not formal in nature. These posts are written, forwarded, texted in an informal manner containing unlimited amount of abbreviations. In order to control this problem of informal style of writing, preliminary processing of these posts becomes an important issue. For example, typos could also be words which can be misplaced and should necessary must be corrected in order to correctly express and evaluate the sentiments attached to them much more accurately and precisely. The following steps have been employed by specifically taking into account the previous research done by the Egyptians in the study conducted by Shoukry in 2012. The steps are:

• Deletion of Universal Resource Locators 

• Being precise in the use of long words in these posts  

• Fixation of the typos 

• Deletion of the prepositions and other stop words. 

In order to verify the proper working of the model, it has been examined upon Facebook. Facebook is an absolutely free of cost sharing and microblogging site which was initially co-founded by Mark Zuckerburg in the year 2005. It has billions of users who have been closely associated with the social media site. It has a lot of characters which are known as posts. Any number of duly registered users could be read as has been investigated by Dinerman in 2010.

In order to test the efficacy and precision of these classifications was done by the active users. Beneath this, the outcomes of the tests have been mentioned.

The usage of Post Archivist has been increasingly done in order to collect various tweets about the main topic. It is actually an analytics engine of Facebook which is primarily used to archive, analyse, and search, save, export and share pots related data.

Various Approaches to Sentiment Analysis

In order to successfully perform this task, an aggregate of about 1000 posts have been collected from the month of January 2018 by primarily using the Posts analytics of Facebook which talked about the three main organisations.

A description of the organization has been provided below:

1. Company P 1: A brand new and a famous satellite TV station in Caribbean that was created in the year 2011
2. Company Q: A bookstore which is originally based in Caribbean and it is considered as one of the main source of books for the world
3. Company R: It is basically a website which provides news which is mainly related with the various local and the regional problems and issues.

When analysing the posts, 400 posts were from Company P, 400 posts from Company Q and 200 posts from Company R (Gautam  & Yadav 2014). The post archivist has subsequently given the posts related details in the MS excel format, which contained the essential user IDs, text and the exact time of the posts.

Four users of Facebook had been asked to gloss over their preferences. They were repeatedly again asked to classify manually each preference and like in three different genres; either positive or negative or neutral. A mean was then done for each single preference to get single classification. The table beneath shows a complete glimpse of the classification:

A miniature and brief application which successfully pulls of the approach on the same preferences and likes has been built and the results shown in the table below shows the main framework of the entire applicability.

Data Set	Organization A	Organization B	Organization C
Positive	203	73	292
Negative	98	70	55
Neutral	99	57	53
Total	400	200	400

Fig 4 Results of automatic annotation

In this portion, a robust comparison of the manual and the automatic segregation outcomes had been done. In this collation exercise, the exactness and advocacy for positive, negative and the neutral segregation had been conducted (Cambria, 2013). Also due to lack of the standards, the manual results had been considered using the following baseline:

Exactness= (Automated results + manual results)/manual results

Recall= (Automated results + manual results)/ automated results

On the basis of the above formulas, the following results had been obtained for each of the three domains that were studied. The table 5, 6, and 7 below shows the exactness and the recall for the classification of posts on the three domains. An average exactness of 70% and an average recall of 66% was obtained in Company P (Nguyen et al., 2013) In Company Q an average precision of 80% and recall of 77% was obtained. Finally, an average exactness of 74% and average recall of 69% was for company R.

Table 9 then showcases the aggregate exactness and recall for the approach with an average exactness of 74.1% and average recall of 74.7%

Fig 5. Exactness and Recall for Company P

Figure 6. Exactness and Recall for Company Q

Figure 7. Exactness and Recall for Company R

Figure 9. Aggregate Exactness and Recall

Conclusion:

Complete understanding the true sentiment of a particular organisation is very important for the decision makers to know and understand what their future actions should be. Very little progress has been done in order to analyse the sentiments in the social media circle. From the above experiments and tests, it is very clear that a close and semantic method could be used, while dealing with the very sentiments to analyse posts can produce a better understanding of the overall image of the particular entity. In future therefore, the following things should be done in connection with the semantic approaches in the case social media.

1. Expanding the Semantic ontology to include various sentiments
2. Expanding the semantic ontology to include other kinds of dialects

• Enhancing the implementation of this approach to showcase more precise and accurate sentiments.

References:

Cambria, E., Schuller, B., Xia, Y., & Havasi, C. (2013). New avenues in opinion mining and sentiment analysis. IEEE Intelligent Systems, 28(2), 15-21.

Gautam G., & Yadav D. (2014). Sentiment Analysis of Facebook Data Using Machine Learning Approaches and Semantic Analysis. Contemporary Computing (IC3). Seventh International Conference

Gillespie, T. (2014). The relevance of algorithms. Media technologies: Essays on communication, materiality, and society, 167.

Kowcika, A., Gupta, A., Sondhi, K., Shivhre, N., & Kumar, R. (2013). Sentiment analysis for social media. International journal of advanced research in computer science and software engineering.

Mohammad, S. M., Kiritchenko, S., & Zhu, X. (2013). NRC-Canada: Building the state-of-the-art in sentiment analysis of tweets. arXiv preprint arXiv:1308.6242.

Newman, T.P., 2017. Tracking the release of IPCC AR5 on Twitter: Users, comments, and sources following the release of the Working Group I Summary for Policymakers. Public Understanding of Science, 26(7), pp.815-825.

Nguyen, T. H., Shirai, K., & Velcin, J. (2015). Sentiment analysis on social media for stock movement prediction. Expert Systems with Applications, 42(24), 9603-9611.c