Presents discussion threads which might be shared by any two nations, we can view the network with every discussion thread exposed as further nodes. We transform the `country-country’ data into `country-thread-country’ information, and after that break the triad into two `country-thread’ dyads. This can be referred to as a bipartite, or 2-mode network (see refs. 20 and 21 for explanations on functioning with 2-mode information). This 2-mode data support us visualise the relationships in between nations or discussion threads, and to recognize considerable structural properties. Sentiment evaluation The content material evaluation is carried out within the MySQL database with custom scripts. Working with the 853 Ser-Phe-Leu-Leu-Arg-Asn web messages located within the network analysis, we execute a sentiment analysis in the messages to recognize the opinions of ecigarettes within the neighborhood. To decide if a message is optimistic or negative, we use a uncomplicated bag-of-wordsChu K-H, et al. BMJ Open 2015;five:e007654. doi:ten.1136bmjopen-2015-model22 of classifying the terms located in every single message. The dictionary of words comes from the Multi-Perspective Query Answering (MPQA) Subjectivity Lexicon (http:mpqa.cs.pitt.edu), which identifies 6451 words as positive or unfavorable, with an extra robust or weak quantifier. In the 853 messages regarding e-cigarettes, you will find over 1.4 million words within the text. For each and every message, we compare each and every word and try to match it against the terms in the MPQA dictionary. When the word just isn’t found, we also apply a stemming algorithm to view if the root word is available. One example is, afflicted is not located within the sentiment list, but we are able to stem the word to afflict, that is identified within the list. In the event the word, or its stemmed root, is found, we apply a score to the message: Sturdy, good = +2 Weak, optimistic = +1 Weak, damaging = -1 Strong, adverse = -2 For the reason that messages may be pretty distinctive in length, the raw scores are inadequate for comparison. Additionally towards the raw scores, we also normalise the scores to handle for message size. We conduct several tests to find out how sentiment may possibly connect with different elements within the network. First, we examine how sentiment scores for ecigarettes examine against topics not connected to ecigarettes working with an independent samples t test. We also use outcomes with the network analysis to seek out PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/21331607 any metrics that may connect country interactions with all the sentiment scores. Results Our final dataset consists of 853 messages posted by members in 37 nations, from July 2005 to April 2012. The number of posts over time might be observed in figure 1. Network analysis Figure two depicts how countries (represented as nodes, or vertices) are linked to one another. A tie connects two nations if they coparticipate in no less than 1 discussion thread (ie, both postmessages in a single thread). The strength in the tie–depicted visually by the thickness on the line–is higher when the two countries share a presence in numerous discussion threads. The size in the node represents degree centrality, or the number of other countries a node is connected to. Inside the 2-mode network (figure three), red nodes represent countries and blue nodes represent discussion threads. Each and every tie now hyperlinks a nation with discussion threads that have been posted by members of that country. Node sizes for every country (ie, red nodes) are reset so they may be each of the similar, but we adjust the discussion threads’ (ie, blue nodes) size primarily based on their betweenness centrality. Betweenness is really a network measure that indicates how frequentl.