Hello everyone. I'm happy to see once again in the first class. As I promised, we will pay special attention to Twitter. Why? Because it is the single most powerful big data source available to social scientists for collecting fine-grained, timestamp data of interaction for local, regional, and global events in addition to individual level. I'm 100 percent sure that you all know what Twitter is. But still I suggest to talk about very briefly. Twitter was launched in October 2006. Falling into the category of social media. It has been conceptualized in many different ways. Twitter has been considered as a microblogging tool, as a short message service, as well as social network. In fact, it is a platform for expressing opinions within 280 characters. Until 2017, it was half as much. By the research conducted on Twitter is so vast, diverse, and is a growing field of study. Well, Twitter is a huge available database with numerous possibilities for scholars. It has a very open application programming interface which makes it an ideal medium for its static. Actually, one of the key differences between Facebook and Twitter is that most of the content on Twitter is publicly accessible via the Twitter API or through resellers such as GNAP and data sift. Whereas most Facebook content is private. Thus, the footprints left by Twitter users, the interactions with the chats can be collected and analyzed. As an example of its richness for data mining, the metadata in each tweet contains not only the text, but also different variables, such as number of followers, favorites, language, geographic location, and so on. In addition, we can say that Twitter has its own technical slang. The first and more and more known characteristic is its brevity and simplicity of 280 characters allowed for posting a tweet. Furthermore, Twitter counts with employment by the users of other elements such as direct replies to tweets replies. The address to other accounts, mentions, and the diffusion of information, retweets. The interactions made by the users shape networks around issue publics under hashtag forming conversations and communities. These hashtag topics, forms themselves, networks of topics which were named issue publics. We come to the question, how can we gather data from Twitter? There are three main methods to gather data through the Twitter APIs. First, Firehose. Second, Rest. Third, Stream. Each of these has different procedures to extract specific data. The first method, Firehose, allows full access to Twitter data without any limitation. The Firehose API provides 100 percent of Twitter data in real-time. Despite the suitability of this method for a Storage, Firehose is not generally used due to it's high monetary costs. Only large companies or institutions with high monetary resources might make use of it. In addition, Firehose is not available directly. That is, it is not public pursue. Only through third party companies such as GNIP and Dataset. Researchers can have access to the fire horse API. Good news are that the other two methods, Rest and Stream are public and relatively easy to access. The possibilities of the rest API Variables are immense. Reading author profiles and followers data, extracting settings, languages, and so on. It also allows with the search API to extract tweets containing specific keywords, words, phrases, or hashtags, geographical boundaries, and user IDs. In comparison to Firehose, it contains however, some limitations. Please pay attention that the rest API has rate limits. A researcher cannot take a full following list of users unless it waits for Tweeter API to provide access every 15 minutes. In addition, the search request can only go back in time for one week, and it only provides a sample of up to one percent of the capacity of the firehose. That is, the API will return at most one percent of all tweets produced on Twitter any given time. This third method to get a tweeter data stream consists in leaving the API call open for a certain period of time collecting data on live. Stream API can be set up to stream tweets with specific keywords, words, phrases or hashtags, geographical boundaries, and user IDs. Like this search API stream provides data up to one percent of the capacity of the firehose. There is no limitation on the time that the coal can be kept open. However, it requires more resources in programming and Infrastructure than search API. Since stream API is a constant open coal to the Twitter API, there is a need to prepare additional coding into programming script in case of connection problems appear during that time the call is open. Usually stream API is opened for several hours, days, weeks, or even months. During the time the call is open, they might be difficulties. This is why additional coding is needed in order to continue data gathering In the case of an internet connection failure. In addition to auxiliary cording stream API requires extensive hard disk space as the data gathered might be large. On average, a million tweets require around one gigabyte of hard disk space. A common solution is to have an external server or server provided by a university to store the collected data. In some, only one methods provides a total amount of data that 100 percent with the firehose API. Well the other two methods collect up to one percent of tweets, depending on the filters imposed by Twitter. This represents a falling problem for the research. This problem is data bias. It means that when the given data is up to one percent of the total amount of data for the event, it is extremely difficult to extract valid conclusions of very large events or global events or of large data sets. The functioning of both public APIs makes it impossible to replicate the same data gathering. This is a major limitation since researchers would be unaware of both the nature of the population they're aiming to analyze and the specific sampling methods used by public API, to satisfy the researchers need of samples. The next limitation of Twitter research is representation bias and it's connected with the weak representation of general public on Twitter for several reasons. First is Twitter small usage compared to other platforms, such as Facebook or news media such as TV, newspaper, or radio. Twitter has around 330 million of active users worldwide, whereas Facebook has 2.6 billion. Second reason is a difference of Twitter usage across the world. In some countries, it is widely used, while in others, its use is marginal. The third reason is the digital divide, which raises additional concerns about generalizing any knowledge from online to offline populations. The Twitter population is different from general population. It tends to be younger and better educated. The next important point about Twitter limitations can be called language challenge produced by language bubbles, which in turn are formed by users who interact in different languages. It is always important to remember about the differences of hashtags in different languages while collecting data otherwise, it can screw the results. To conclude, there are some methodological caveats which needs to be taken into consideration while conducting research on Twitter data. However, even if the Twitter world is not identical to the offline world, it is entirely real because its users, their desires, emotions, and political attitudes are real. So why is Twitter so important for the international relations researcher? Well, first, because of its unique degree of transnational communication and the open interactivity among users. Just think about it. It makes the platform an ideal public arena within principle 'no restrictions'. Second, due to its asymmetric and open principle of "following" users without mandatory reciprocity. We can track which states, politicians, and international organizations follow each other and which do not. To conclude, at this stage Twitter with its very open APIs is an ideal medium for study, since it gives us an opportunity for dynamic analysis of political data.
