Table 1.

Dataset characteristics

Dataset name	MedNLP	Dummy-EHRs	Pathology Reports
# of documents	50 reports	32 pairs of records and summaries	1000 reports
# of sentences	2244	8183	3012
# of tokens	42,621	154,132	194,449
# of all tags	490	3017	295
# of age tags	56	39	0
# of hospital tags	75	170	31
# of person tags	0	135	224
# of sex tags	4	16	0
# of time tags	355	2657	40
Example in original Japanese text	工場に勤めている<a > 64歳</a > の < x > 男性</x > 。	施設入所中で寝たきりの<a > 86歳</a > <x > 女性</x > 。全介助	<<院外標本 <h > 静大皮フ科クリニック</h > 、 < p > 桑田 智</p>
Example translated into English	A < a > 64-year-old</a > <x > man</x > works in a factory	An <a > 86-year-old</a > <x > woman</x > bedridden in a nursing home. Total assistance required	<<Ex-hospital sample < h > Shizudai Dermatology Clinic</h > , < p > Satoshi Kuwata</p>