![]() ![]() In this article, we present a dataset containing history-related content obtained from social media. ![]() In the same year (1916) another remarkable event occurred -the British Army suffered its worst day with the loss of 19,240 men in the Battle of Somme (hence, the hashtag. This is because 2016 marked the 100th anniversary of the Battle of Ver- dun which is especially remembered due to its estimated nearly 1 million casualties. 3 shows that there are many mentions of 2016 with #ww1. We can notice thave strong connection with hashtags #ww1 and #wwii as the two events were held during these respective years. shows the most common hashtags used with content con- taining the peak years of Fig. In this work we employ AIDA -an an- notation tool linking phrases in short text with their corresponding Wikipedia articles -for detecting entities. 4 which lists the top entities in our dataset. 2 that shows the most common entities corresponding to the peaks and at Fig. This can be confirmed when looking at Fig. ![]() WWII started in fact earlier with the Nazi invasion on Poland in 1939, many history-related tweets in our dataset originate from USA and Canada due to the chosen English hashtags resulting in the focus on the North American involvement in the war. As Hei- deltime adds for such expressions "00" at the head of year information (e.g., 0016, 0088) we converted "00" to "19" or to "20" depending on whether the last two digits are less (conversion to 20) or more than 17 (conversion to 19). While 5 Note that some tweets contain abbreviated temporal expressions (e.g., 6/11/16 or 3/19/88). Two dates common for WWII are: 1941 denoting the Pearl Harbor attack and the subsequent participation of USA in the war, and 1945 which is related to the Normandy landing and the end of the war. 1 which represent two key events in the last century: WWI and WWII, and year 2016. The formal definition of the probability distribution for a year y is given in. We then combined for every year all the computed probability distributions based on all the tweets in our dataset. In other words, for a given time reference (e.g., 1960s) with t b denoting its start year (1960) and t e indicating its end year (1969) we set the probability distribution with zero values for t t e (e.g., before 1960 and after 1969) and with non-zero values for t b ≤ t ≤ t e that sum to 1 (e.g., 1/10 for each year from 1960 to 1969). To plot such a curve, we converted the extracted temporal references to probability distributions over their corresponding timespans using year level granularity. 1, the remembering curve as it reflects the strength of the collective attention of users towards different time periods of history. We map all the extracted temporal expressions on timeline as shown in Fig. first analyze which time periods users are interested in by investigating time references included in tweets. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |