Article Preview
Top1. Introduction
Text summarization plays a significant role in information retrieval and data processing systems. According to Forbes, an average of 2.5 Quintillion bytes of data are produced every day and 90% of these data growth has been seen in the last two years (Dalwadi, 2017). It becomes necessary to store this data in a structured format and analyze for business needs. The methods like PageRank of information retrieval based on graph data structure can be used for text summarization (Shams B., 2018). It is always a challenge for the data science community to analyze, evaluate and find actionable insights from a large dataset. Handling such huge data and extracting the meaning from it in a short time such that it gives maximum crust is a challenging job. The extractive text summarization can be used to analyze unstructured data extracted from social media sites, reviews and feedback systems. It applies methods like stemming, lemmatization, and feature extraction through Tokenization and Vectorization (Singh J., 2018. Machine learning techniques based on text summarization methods are discussed in (Allahyari M., Dalwadi, B., 2017).
Figure 1. Architecture of recommender systems
A pre-analysis is always crucial for the decisions on a release date, advertising strategy, promotion techniques. As per (Singh J., 2018) promotional strategy of movies is based on how the public is reacting on trailer launches. Twitter also provides a universal platform to everyone for sharing their thoughts in the form of tweets. Twitter handlers provide tweets related to a particular keyword of a movie like a title, actor, director, etc. As per (Fang, C., 2017), each comment and tweet can be considered as a sentence. and various sentence scoring algorithms applied for text summarization to identify representative sentences. Here, an algorithm such as CoRank and CoRank+ can be used for word-sentence scoring techniques. Also, there are several algorithms present and it has been seen that they show different behavior on the various dataset (Ferreira, R., 2013, Gupta, D. P. N., 2012, Singh, D. P., 2013). Movie RS shows in Figure 1 extract data from different websites like Netflix, YouTube, scene unseen, etc. These texts are posted by different users through various web and mobile applications. Click streams for watching different movie trailers and advertisements also become a crucial data source for RS (Xu, Y., Zhang, F., 2019).
1.1. Background
Many online movie streaming sites usages, Recommendation System (RS) to recommend movies to their potential users. As per (Wei, S., Zheng, X., 2016) in RS the prime problem is to deal with a cold start since not all the users are ready to review and rate the content they watch. Da Costa et al. (da Costa, A. F., 2019) presents RS like EcoRec, a hybrid approach of two or more algorithms that takes a single input dataset, combines the output of both into a single ensemble. RS analyses the emotive response and social media trends of the movie in the crowd to make major decisions for business profits as discussed by Verma et al. (Verma, J. P., 2019). RS takes into account the pattern of user browsing and recommends users with similar items. RS suggests similar things whenever you want to select a product through online shopping (Kumar, S., 2012). Mangold et al. (Mangold, W. G., 2009) shows that social media has given a platform to many firms for communicating with their customers. For instance, if we look at old promotional activities, social media marketing always beats them, for example, direct mails. Previously, promoters used to send promotional emails to users and it was kind of hit until spam classification was invented. According to TextRequest (Miller, R., 2019), users receive an average of 88 emails per day, while they send 34 emails per day. But the study also says that about 49.7% of these emails are classified as spam and go directly into spam folders which either are deleted or never read.
Figure 2. Leading country based on number of monthly active Youtube users as of first quarter 2016
Figure 3. Number of YouTube users worldwide from 2015 to 2021 in millions, (c) Number of Twitter users worldwide from 2014 to 2020 in millions
Figure 4. Number of Twitter users worldwide from 2014 to 2020 in millions
Figure 5. Online Statistics -Online buyers vs years
Figure 6. Online Statistics - E-Commerce Sales growth rate
Figure 7. Online Statistics - Sales in billion US dollar vs. Year