Table 4:
F-scores [%]. MT-All (shown in bold) generally outperformed the ST models. “OS” denotes oracle span setting. “Macro” represents macro-averaged F-score. * shows statistical significance p < 0.05.
![F-scores [%]. MT-All (shown in bold) generally outperformed the ST models. “OS” denotes oracle span setting. “Macro” represents macro-averaged F-score. * shows statistical significance p < 0.05.](https://mitp.silverchair-cdn.com/mitp/content_public/journal/tacl/10/10.1162_tacl_a_00481/1/m_tacl_a_00481_i004.png?Expires=1682941251&Signature=qTlJR1OouoyTqfanoFhLsDY8MJlaerz9pKtsvxlcv8v2b4QaYJuYmMBh1tGc4I6oPLmrJTFyOGNoewuL0CDZzXERWfF26R~9IPzsZGV30UHW5HsHeFyU76ZLpEMjFgw~5AdBtyAXvWvlfkVvudDJXOoHNyA3ZpguavnUbIklt6UmfGRf3-aGas8-lnzOLQLy6e-neN8XTZ~eMgPIqIfwf6VkINBR9NXGtuzz0-S2uK8z8px7Eq3aJ9NBmqWcYK2Hfb7nkB613YD~qK95K~L92j9giHtlGcvpv9ewwuE35dTM8sVhl4pVbvjFxw~y08QcuGjK40WdWQOYld8uyli4Zw__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA)
F-scores [%]. MT-All (shown in bold) generally outperformed the ST models. “OS” denotes oracle span setting. “Macro” represents macro-averaged F-score. * shows statistical significance p < 0.05.