Social media buzz created by #nanotechnology: insights from Twitter analytics

1 Introduction

There has been a lot of buzz recently about the word “nanotechnology” in various forms of media, be it print media, e-media or social media. Social media has seen an unparalleled growth as compared to other forms of media in recent years as it enables people from different nations having common interest to communicate on and discuss various topics [1]. Even scientists and researchers have become a part of this virtual world, where they share their current research with the global audience. One such area of research where scientists from all around the world are contemplating to produce the novel applications is nanotechnology.

Nanotechnology is a technique where high circuit integration density can be realized on a single chip by scaling down the size of electronic devices. The scaling down of electronic devices is driving electronics into the realm of nanoelectronics. It is being professed that Moore’s law could cease to exist after 2020 [2]. Faster response and integration into small area are the major gains of molecular electronic devices when compared to its silicon counterpart [3]. Deoxyribonucleic acid (DNA)-based molecular wires are fast emerging as favorable candidates for molecular electronics. The self-replication property of DNA advocates its fitness for the design of alike molecular electronic devices. The research work of Bhat et al. [4] illustrated the choice of best electrode material in the three-dimensional (3D) nanostructure of adenine, which is one strand of DNA. The role of crystallographic orientation of electrodes has also been determined to see the transport properties through DNA [5]. Thus, over the years, various research works on conductivity through DNA have been carried, and these works can be helpful to produce an insight toward DNA-based molecular devices like photonic logic gates, memories and transistors.

Similarly, ever since the knowledge of mankind expanded, only two allotropic forms of carbon were known, namely, graphite and diamond, but the discovery of fullerenes by Harold W. Kroto and his team in 1985 [6] revolutionized the perception of scientists, and carbon emerged as wonder element of nanoage. The fullerenes, generally present in the form of spherical balls, can be constructed by pentagon and hexagon rings of N carbon atoms. However, only two isomers have been experimentally realized, C60 [6] and C20 [7] fullerenes.

The basic structure of such fullerene molecules is the thick planar sheet of carbon atoms generally arranged in honeycomb pattern, called graphene sheet. This graphene sheet was discovered in 2004 when Andre Geim and Kostya Novoselov used a specialized technique called micromechanical cleavage and extracted a single sheet of graphene from a 3D graphite [8]. Such sheets can be molded into a number of isomers, namely fullerenes, carbon nanotubes, etc. Various contemplations have been made to understand its properties for multidisciplinary applications, be it medical [9], chemical, physical, electronic and other interdisciplinary fields [10].

The aim of this paper is to critically analyze the social media hype created by the word “nanotechnology” and related keywords by monitoring and applying latest data processing strategies on the Twitter traffic.

2 Literature review

Science and mass media have always shared a strong bond. Whether print media or social media, both have played a crucial role in mapping public opinion in regard to new technology. Earlier, it was newspapers and magazines that used to cover the latest trends in technology and tried to map public opinion towards these. But in the last decade things, have changed and researchers have opted for social media in addition to the traditional print media.

This section is categorized into two parts: (a) scientific discussion in various media and (b) nanotechnology in media. The first part focuses on various media platforms used for communication about new technologies, while the later discusses various work carried out with regard to “nanotechnology” in mass media.

3 Research methodology

To accomplish our objectives, we used a very simple yet realistic technique. This technique starts with data (tweets) collection from a famous social media website, i.e. Twitter. Tweets were collected in a time-specific manner starting from January 1, 2018, to January 31, 2018. Since the collected tweets contained a lot of unwanted material, they were preprocessed. Once preprocessing was done, we applied various social media analytics technique ranging from tweet statistics, (#) hashtag analysis, geo-location analysis, and sentiment analysis. The detailed description of each task is explained in the upcoming sections.

4 Data collection

For collection of tweets in a trusted manner, we developed a system using ASP.NET [39] and further integrated tweetinvi API [40] to fetch tweets in json format from Twitter. Tweets were collected on a daily basis based on specific (#) hashtags. For selection of these hashtags, two independent expert teams (nanotechnology and allied fields) were constituted having three members each. Both teams gave keywords that were directly or indirectly related to nanotechnology, like “nanotechnology,” “nano, “nanoscience” etc. These keywords were used as (#) hashtags to search nanotechnology-related tweets from Twitter.

An example of tweets fetched based upon these specific (#) hashtags is shown in Figure 1. The sample tweet shown in Figure 1 contains three hashtags (#graphenepower, #nano and #nanotechnologies) that matched the keywords given by the expert team.

Figure 1:

A sample tweet.

A total of 6289 tweets were collected in 31 days starting from January 1, 2018, to January 31, 2018. The daily tweet collection is shown in Table 1. The collected tweets contained tweet message, tweet sender, date and time and location (country) from where the tweet originated. Note that the location of all the tweets was not available due to security reason and free API policy.

5 Detecting irrelevant tweets

Since many keywords suggested by experts were ambiguous, i.e. they were related to multiple other entities apart from nanotechnology, this resulted in fetching of various irrelevant tweets that were not relevant to our context of analysis. Hence, it was necessary to remove them.

For this, the same expert teams (see Section 4) were asked to manually investigate the collected tweets. After manual investigation, it was identified that some tweets were not relevant to our context of analysis. Two authors performed the operation of removal of these irrelevant tweets. Examples of irrelevant and relevant tweets are shown in Figures 2 and 3, respectively. A total of 754 such irrelevant tweets were removed and we were left with 5535 relevant tweets.

Figure 2:

A sample tweet (irrelevant).

Figure 3:

A sample tweet (relevant).

Since on Twitter, a person can post tweets multiple times, this makes our collected data biased. This could lead to a scenario where the results of our analysis could be diverted toward the opinion of people posting multiple tweets. So, in order to perform a fair analysis, we applied the same technique used by Singh et al. [41] for removing multiple tweets from the same person. This technique identifies all tweets that are from same users and marks them, leaving only the first tweet. This technique also helps to neutralize the effects of bots, which create unnecessary tweet traffic. After applying this, we were left with 3671 tweets. Note that this process is not a substitution of normalization [42], [43] or bot detection [44] but helps to perform a fair analysis.

6 Data preprocessing

Since the collected tweets contained a lot of noise and unwanted stuff, it was necessary to preprocess them before applying any analysis operation as this unwanted stuff may lead to undesired results [45]. The preprocessing operation includes removal of punctuations, numbers, English stop words, web links and extra white spaces present in the tweets. The entire task of preprocessing was performed using R-language [46]. Table 2 shows the details of the preprocessing task.

7 Social media analytics

Social media analytics is an art of extracting important hidden information from user-generated content on social media and using this information for decision making [47], [48]. In this section, various social media analytics techniques are applied to extract important information regarding nanotechnology from the preprocessed tweets.

7.2 (#) Hashtag analysis

Hashtag analysis [50] deals with various hashtag statistics. A total of 14,366 hashtags were identified in 3671 tweets, out of which 2743 hashtags were unique. A total of 5386 tweets contained more than one hashtag. The hashtag with maximum occurrences was #nanotechnology (1733). The top five hashtags with most occurrences are shown in Figure 4.

Figure 4:

Top five (#) hashtags.

7.3 Geo-location analysis

This analysis takes into consideration the locations from where the tweet originated [51]. However, due to privacy reasons, only the location of those tweets is returned wherein privacy settings allow us to do so. Out of 3671 tweets, 514 were geo-location enabled. Out of these 514, maximum tweets originated from the US (141), closely followed by India (107). The result of the geo-location analysis is shown in Figure 5.

Figure 5:

Results of geo-location analysis.

7.4 Sentiment analysis

Sentiment analysis is a text analysis technique that deals with the extraction of sentiment from a given piece of text [52], [53]. Sentiment analysis further involves two suboperations: (a) polarity analysis [54] and (b) E-motion analysis [55].

1. Polarity analysis: This deals with identifying the polarity (i.e. positive or negative) of a given text. We used WEKA tool [56] to perform the polarity analysis. A classification model was built using a supervised machine learning algorithm, support vector machine (SVM) [57], to classify the given tweets into positive or negative tweets. The reason for choosing SVM over other algorithms was that multiple studies have concluded that the SVM gives best results in sentiment analysis [58], [59], since supervised machine learning involves training and then testing of the dataset. For training purposes, we have used IMDb [60] dataset consisting of 500 positive and 500 negative instances. The reason for choosing the IMDb dataset for training was that it is often considered one of the best and authenticated datasets for supervised machine learning [58]. Once training was completed, we tested the classifier on preprocessed tweets. The results of polarity analysis are shown in Figure 6, which indicates a highly positive polarity among tweet senders.

2. E-motion analysis: This deals with categorizing data based upon eight e-motions. E-motion analysis was performed using R-language. The results of the e-motion analysis are shown in Figure 7. These results clearly indicate that the tweets contained more positive words (anticipation + joy + surprise + trust).

Figure 6:

Results of polarity analysis.

Figure 7:

Results of E-motion analysis.

8 Discussion

Social media represents a virtual world where people around the world can have a two-way communication on various topics, including those related to science and technology. This paper aims to explore one such booming technology: nanotechnology. The results of our research are quite encouraging as the results show an overall positive attitude among people who are involved in this conversation. Most of the discussion focuses upon various applications of nanotechnology, be it in medical field for cancer detection or the various optical application of graphene. This was also a strong reason that graphene was one of the most commonly used hashtags (#) as depicted by our study.

However, there were few challenges that are needed to be addressed. The data collection period of 1 month was too short; had this analysis been covered over a period of 6 months or so, it would have delivered more useful results. Further, we can detect what events lead to polarization of public opinion. Another important challenge that was not addressed in this study was the normalization of tweets [42], [43]. Twitter has highly varying usage rates per person for different geographical regions. This depends on people’s privacy concerns, available infrastructure, age of the population etc. Normalization of tweets can certainly help in answering some of these issues. Similarly, bot detection [44] remained another important challenge that was not addressed.

All these motivate us to further extend this study to a larger time frame, which can produce even better results by applying other social media analysis techniques like topic modeling and community analysis and addressing challenges that were not addressed in this study.

9 Conclusions

With the rise in social media users, social media analytics have become an important tool. In this paper, we used social media analytics to analyze the buzz created by nanotechnology on Twitter. Our results indicate that there is a lot of talking going on in Twitter regarding nanotechnology and in a positive sense, too. Our results clearly indicate that there is lot of positive sentiments attached to the subject nanotechnology as trust in the technology, anticipation about the outcome and joy about the availability of the predicted applications overweigh many folds the anger due to nanowaste, mistrust about the failures in achieving the target and anger of wasting tons of taxpayers’ money in projects related to nanotechnology.

However, tweets collected in a 1-month time period are not sufficient to analyze such a vast topic, and these results could be still improved by extending the tweet collection period. We can substantiate this research topic using a more suitable dataset by comparing nanotechnology with other booming technologies like artificial intelligence and internet of things.