Read prosecutor selection alongside the speech and its evidence. · 将控告者选择与演说、证据并读。
The whole speech beside the translation the reader carries, paired sentence by sentence: 320 pairs over 334 Latin sentences and 400 English ones. Every pair says how it was arrived at, and the ones that should not be leant on say so.全篇与读本所用译文逐句并列:拉丁 334 句、英文 400 句,共成 320 对。每一对皆注明其所由来,其中不足凭信者亦一一标出。
pairs对
320
one Latin sentence to one English一句对一句
243
marked uncertain标为存疑
107
sections whose English begins mid-sentence英译始于句中之节
33
The translator's sections are not the speech's. In 33 of the 73 sections the English begins in the middle of a sentence, because the clause it renders was carried back into the section before. Pairing the two section by section therefore fails at every one of those boundaries, and the alignment here is computed over the whole speech at once, so that the boundary does not exist.译者的分节与原文的分节不同。七十三节之中有 33 节,英译起于句中,因为所译之句已被带入上一节。故逐节比对必于每一处交界失效;此处的对照乃就全篇一次求得,交界之患遂不复存在。
Each candidate pairing is scored on three things: how many of the English words the Latin sentence's own vocabulary predicts — every Latin word is put through this site's morphology table to its headword and through the gloss table to its English — how many of its proper names appear, and whether the two are of a length that a translation would produce. The best monotonic path over the whole speech is then taken, allowing a Latin sentence to answer to one, two or three English ones and the reverse.每一种可能的配对按三事计分:拉丁句自身的词汇所应产生的英文词有多少见于英文句(每个拉丁词先经本站形态表归至词目,再经释义表得其英文);其专名有多少出现;以及两者长短是否合乎译文之常。然后就全篇取最佳的单调路径,并容一句拉丁对一句、二句或三句英文,反之亦然。
243 pairs are one Latin sentence to one English and can be read straight. 107 are marked uncertain: every pairing that merges sentences, and the lowest-scoring eighth of the rest. A pairing marked uncertain is still the best available reading of where the translation of that sentence lies; it is not a claim that the two say the same thing.其中 243 对为一句对一句,可径读。另有 107 对标为存疑:凡合并数句者,以及其余之中得分最低的八分之一。标为存疑者,仍是就该句译文所在所能给出的最好判断,并非断言两者所言全同。
The pairing is checked by something it was not built on: a proper name in the Latin should turn up in the English beside it. Of 88 pairs that carry a name, 2 do not show it, and both were read: in each the pairing is right and the translator has compressed. They are listed on the pairs themselves. The check runs on every build, and a pairing that loses a name without being either marked uncertain or recorded here stops the build.此对照另以一事相验,而此事非其所本:拉丁句中的专名,理应见于与之相配的英文。载有专名者 88 对,其中 2 对不见其名,二者皆经披阅:配对无误,乃译者删削所致,已就该对注明。此项查验每次生成皆行,凡失其名而既未标存疑、又未著录者,生成即止。