A Case Study of Analyzing Facebook Posts of Different Bangladeshi Newspapers

Introduction

Media in the present globalized world has not been the same as it was when people had to wait for the morning papers or radio news to follow various occasions around the universe. Likewise, news creation, spread, and utilization are no longer same as the pre-innovation era when people only relied on twenty-four-hours news media. However, presently, a increasing number of people seeking news are continually going online to know about the most updated incidents around the globe. Social media is rapidly changing the media scene in the recent condition of news creation and spread.

Nowadays, with the quick pace of globalization influencing all aspects of life and upsetting the data innovation which has fundamentally altered the procedure of news generation and its dispersal. Currently, all the users of internet can without much of a hassle engage themselves with a global stage where news is openly accessible and easily spreading out to others. To define the term ‘social media’ in accordance with Wikipedia, Wikipedia refers that social media are online-based innovations that allow the users to share information, thoughts, interests and different and different types of articulation through virtual networks. Facebook, LinkedIn and Twitter are some of the popular social networking and communicating sites that allow their users to keep in touch with the sites through updating text, picture, video or status. It is undeniable that the convention standard of media practices and journalism are getting altered due to the interactive aspect and significant role of social media in communication and breaking news.

The news social media websites provide is nearly in a textual format which is considered as unstructured text. Text mining is particularly used as the way of bringing out indefinite and practical models or information from a summation of enormous and unstructured data or corpus. Computational linguistics, Information Retrieval (IR) and data mining are some research fields that are being incorporated by one of the branches of data mining named text mining, scholars stated. A good number of text mining techniques such as ‘topic detection and tracking, keyword extraction, sentiment analysis, document clustering, and automatic document summarization’ have been introduced to increase the efficiency of analyzing text. Besides, NPL is correlated field with text mining which has a concern about the mutual relations between vast amount of unstructured texts that exist. News generation and consumption of User Generated Content has made the recent media discussion, a possible field of research.

Motivation

Bangla has become one of the most extensively spoken languages around the world. A good number of Bangla posts are being shared by the Facebook pages of different Bangladeshi newspapers on a daily basis. The texts of Bangla language are unstructured which need to be transformed into informative knowledge from a vast amount of data through applying various text mining techniques. Due to having lack of literature on the analysis of Bangla text, to be more precise on Bangla news, the present study seeks to explore the model of analyzing newspapers textual Bangla news exist on social media. The key reason behind getting started of my research is the availability of vast amount of Bangla texts that I want to transform into constructive knowledge.

Background

The motive of having background study is to assist the present research being studied. It is an initial research stage to design a thesis as there are many issues to be apparent and thoughts to be clarified. All the relevant existed works have been done through different text mining techniques belong to this section along with the scopes and challenges.

An increasing number of readers and writers are being attracted successively by social media, several scholars disclosed that. The influencing part of a wide range of web-based social networking lies in its engaging quality and overall reputation within the internet users around the global village. Due to the global move of social media for example Facebook, mass media has been outdated now, it is all about personal media today. It is quite undeniable that social networking site like Facebook has one of the most popular sources of contemporary news where users have access to the Facebook pages of newspapers and they have option to choose what to read or not. Basically, what we, online users do when we go online is selecting kind of news or views we really care most.

An investigation reviewed different techniques of text mining to analyze textual model of social network along with online-based applications. A survey revealed that authors targeted to provides vast idea of different text mining techniques and their exercises in social sites. Classification and clustering are two of the recently developed key approaches to text mining in the aspect of intellectual unstructured text analysis.

A recent study has been done on text mining and analytics, a case study that analyzed unstructured English text of different news channels Facebook posts. The study showed several techniques of analyzing ambiguous raw data sets and their transformation into quantifiable data. Research relied on a built tool used for gathering Facebook data and analysis process was performed by RapidMiner, an integrated environment for data science operations.

Paying particular attention to “Arab Spring”, examined during this vital phase of history, Facebook intending to collect convenient details about online users' sentiments. Based on Support Vector Machine (SVM) and Naïve Bayes the analysts utilized a system to this end. Besides, a lexical resource for sentiment analysis is formed which is extracted from the emoticons, interjections and acronyms derived from the updates for users’ statuses. Although the research achieved profound findings about Tunisian Facebook users' In January 2011, Tunisian revolution which is one of their recorded moments, it represented a few flaws identified with targeted users' changing emotions on a specific point. The investigation disregarded the factor of time reliance in its examination and exchange which influenced the findings incompletely. If the study incorporated the time-related feature in its investigation, it would have been more fascinating.

A little research has been directed to investigate the enormous information posted by customers on a daily basis which can be fruitful for organizations' benefits. However, the use of social networking sites has moderately been increased day by day. To disclose how the analysis of social media data can become significant to the decision makers and management research and practice, seek to present a case study. Data was collected through the SAMSUNG mobile Facebook page. By using ‘NCapture for NVivo 10’ 128,371 comments were captured that represented the corpus of research from 10th June to 10th September.

The structured approach has been suggested by the researchers to analyze social media data that include only the comments in English language. To extract quantifiable data from social media, researchers outlined a straightway approach to existing knowledge. The consequence of such quantification can be performed in studies, surveys and the plan of decision-making frameworks. Yet, the research failed to notice regularly changing example and progression of Facebook users.

As social media platforms allow immense space for anyone to express and exchange their opinions, feelings and views, students are considered to be identical either. Therefore, this area was a prolific field of study for various researchers. Researchers considered students casual discussions via web-based networking media concentrating on their emotions, opinions and worries about their learning knowledge. An example of 25.000 engineering students’ tweets related with their school life was examined by the researcher. The consequence of the investigation uncovered that various problems for example study load, sleep agony and lack of social engagement.

Moreover, an investigation was focused on extracting knowledge from university students’ information available on social media sites. Using K-means, a data mining technique to extract constructive information of educational sector, the author conducted a questionnaire for university students from different field of studies and analyzed the answers through data mining technique. Facebook, Orkut, and Twitter are most frequent sites used by the university students, study revealed.

Research Summary

Text mining, learning technologies and analytics have been fruitful for the scholars and novice researchers interested in gaining practical experience. The drawbacks regarding laboring analysis of qualitative data and user-generated textual contents from a vast amount of data have been overcome.

Scope of the Problem

One of the vital aspects of study which is newspapers’ social networking data analysis seems to be overlooked, though significant research on social media data mining has already been performed. To be more precise, no research has been recorded regarding unstructured Bangla news analysis yet. Therefore, current study seeks to have meaningful information after analyzing a large scale of data sets extracted from three popular newspapers’ Facebook pages.

Challenges

Despite having the availability of a huge amount of the data over the internet, extracting a good number data had been a challenging part.

One of the most challenging parts I had been though faced throughout the research was having compatibility of Bangla text with the existing system. Besides, Bangla language contains more stop-words along with different punctuations and digits compared to English Language. Moreover, preprocessing phase became more complicated as extracted data from Facebook were full of irrelevant variables that demanded to be eliminated in order to have the efficiency in data sets.