Precision vs Recall: author's mnemonic memo via オオカミ少年 and praecidere

The memo opens with the author's own experience: the names of ML evaluation metrics were hard to remember as a student, and only after entering the workforce did they become common in conversation, prompting this summary. The first mnemonic pairs Type I error with FP (false positive, wrongly flagging positive) and Type II error with NP (false negative, wrongly flagging negative). Rather than etymology, the author settles on the 'オオカミ少年' story: the boy first lies that a wolf appeared when it did not (false positive, Type I), then later a real wolf is treated as not detected (false negative, Type II). The author adds a caveat that strictly the tale is about a true positive being ignored rather than a false negative. For Precision and Recall, the memo contrasts the English terms: Precision from Latin praecidere ('to cut off / trim'), implying trimmed purity, or the share of predicted positives not contaminated by FP; Recall as 'calling back / remembering', the share of actual positives retrieved, falling as FN increases. The Japanese terms 適合率 and 再現率 are also explained, with 適合率 linked to information retrieval's idea of results fitting the user's request, and 再現率 described as nearly the same as Recall. The author notes the memo will be updated as needed and invites corrections.