Table 3 Statistics of the TKG datasets (\(\mathscr {D}_{train}\), \(\mathscr {D}_{valid}\), and \(\mathscr {D}_{test}\) are the numbers of facts in training, validation, and test sets; \(\Delta \textrm{t}\) represents time interval).
Dataset | ICEWS14 | ICEWS05-15 | ICEWS18 | GDELT |
---|---|---|---|---|
\(\left| \mathscr {N}\right|\) | 6,869 | 23,033 | 10,094 | 7,691 |
\(\left| \mathscr {R}\right|\) | 230 | 251 | 256 | 240 |
\(\mathscr {D}_{train}\) | 74,845 | 368,868 | 373,018 | 1,734,399 |
\(\mathscr {D}_{valid}\) | 8,541 | 46,302 | 45,995 | 238,765 |
\(\mathscr {D}_{test}\) | 7,371 | 46,159 | 49,545 | 305,241 |
\(\Delta \textrm{t}\) | 24 hours | 24 hours | 24 hours | 15 mins |
\(\left| \mathscr {T}\right|\) | 365 | 4,017 | 365 | 2,975 |