As Twitter shrinks, the percentage of the remaining tweets that include verified geographic information has plunged as well, meaning a greater and greater loss of geography.
When all tweets were taken into consideration, including those that did not mention the CHEO case, 38.7 % (n = 120) opposed the patenting of genes and the remaining tweets (n = 190) were neutral in tone.
Since many tweets lacked geographic information, this model was trained and tested on significantly less data (905,497 tweets for which we could accurately infer the US state of origin, less 90,000 of these belonging to Region 2); the remaining tweets were excluded from this analysis.
The labeled tweets were used to train a classifier which then automatically labeled all the remaining tweets.
This trend continued also after the referendum day (June 24), when the share of Leave tweets was 23.6% and the share of Remain tweets was 22.7%.
Of those remaining Place tweets, 98.7% specify only the user's city, state or country.
More precisely, the StanceScore of a user is computed from the number of the Leave tweets L, the number of the Neutral tweets N, and the number of the Remain tweets R, as follows: begin{aligned} {StanceScore} = frac{R-L}{R+N+L} end{aligned}The formula is derived from the sentiment score [32] which is computed as the mean of a discrete probability distribution of sentiment labeled tweets.
The basis of their prediction is the percentage of Leave-related tweets versus the number of Remain-related tweets.
If the user is balanced in the number of Leave and Remain tweets, or prevailingly neutral, her/his stance it assigned to be Neutral.
Porter stemming of all remaining terms in tweet.
