「#3889. ネイティブがよく間違えるスペリング」 ([2019-12-20-1]) でネイティブの犯しがちなスペリングミスの例を挙げた.1993年の NFER (The National Foundation for Educational Research) による調査では,11--15歳のネイティブの子供が犯しやすいスペリングミスが明らかにされた.NFER はスペリングミスを以下の5種類に分類している(以下の引用と数値は Cook (124) より).
・ insertion of extra letters, such as the <l> added to 'until'
・ omission of letters, such as the <r> missing from 'occurring'
・ substitution of different letters, such as <a> instead of <i> in 'definate'
・ transposition of two letters, such as <ei> for <ie> in 'freind'
・ grapheme substitution involving more than two letters but only a single cause, for example when an equivalent according to sound correspondence rules is substituted for the usual form, as in 'thort' for 'thought'
5種類のミスの内訳をみてみると,insertion (17%), omission (36%), substitution (19%), transposition (5%), grapheme substitution (19%), その他 (3%) ということである.この数値を見ておよそそんなところだろうという印象だったが,grapheme substitution が予想よりも多かったので,ちょっとした発見だった.成人の数値はまた異なってくるかもしれないし,さらにネイティヴではなくL2学習者ならば,やはり別の数値が出るのではないかと疑われる.
もちろんL1, L2学習者にかかわらず,このスペリングミスの5種類の分類は単純明快で,一般的に利用できるように思われる.スペリングのミスのみならず,その変化や変異の類型にも役立つ分類だろう.
なお,スペリングの差異を測定する方法としては,ほかに "Levenshtein distance" (levenshtein_distance) というものもある.これについては「#3406. Levenshtein distance」 ([2018-08-24-1]),「#3399. 綴字の類似度計算機」 ([2018-08-17-1]),「#3397. 後期中英語期の through のワースト綴字」 ([2018-08-15-1]),「#3398. 中英語期の such のワースト綴字」 ([2018-08-16-1]) を参照.
・ Cook, Vivian. The English Writing System. London: Hodder Education, 2004.
文字列の類似度や相違度を測る有名な指標の1つに,標題の "Levenshtein distance" というものがある,「#3397. 後期中英語期の through のワースト綴字」 ([2018-08-15-1]),「#3398. 中英語期の such のワースト綴字」 ([2018-08-16-1]),「#3399. 綴字の類似度計算機」 ([2018-08-17-1]) でも前提としてきた指標であり,「#1163. オンライン語彙データベース DICT.ORG」 ([2012-07-03-1]) でもこの指標を利用して類似綴字語を取り出すオプションがある.
考え方は難しくない.もとの文字列から目標の文字列に変換するには,いくつの編集工程(挿入,削除,置換)が必要かを数えればよい.例えば,kitten から sitting へ変換するには,kitten →(k を s に置換)→ sitten →(e を i に置換)→ sittin →(g を挿入)→ sitting という3工程を踏む必要があるので,両綴字間の Levenshtein distance は3ということになる.通常は挿入,削除,置換の編集工程にそれぞれ1の値を割り当てるが,各々に異なる値を与える計算の仕方もある.
以下に,通常の重みづけで Levenshtein distance を計測する CGI を置いておく.試しに「#1720. Shakespeare の綴り方」 ([2014-01-11-1]) より25種類の異綴字を取り出して,カンマ区切りなどで下欄に入力してみてください(要するに以下をコピペ).Shakespeare, Schaksp, Shackespeare, Shackespere, Shackspeare, Shackspere, Shagspere, Shakespe, Shakespear, Shake-speare, Shakespere, Shakespheare, Shakp, Shakspe?, Shakspear, Shakspeare, Shak-speare, Shaksper, Shakspere, Shaxberd, Shaxpeare, Shaxper, Shaxpere, Shaxspere, Shexpere
綴字間の類似度や相違度の計測は,曖昧検索やスペリングチェックなどの実用的な目的にも応用されている.標準的な綴字がなかった古い英語の研究にも,ときに役立つことがありそうだ.
この2日間の記事「#3397. 後期中英語期の through のワースト綴字」 ([2018-08-15-1]),「#3398. 中英語期の such のワースト綴字」 ([2018-08-16-1]) で,異綴字間の類似性を計算するスクリプトを利用して,through と such の様々な綴字を比較した.このスクリプトは,ある程度使い勝手があるかもしれないと思い,より汎用的な形で CGI を組んでみた.
ところが,スクリプトの内部的な仕様の関係でサーバ上で動かないということが発覚.残念無念.公開しても無意味であることを承知のうえ,以下に置いておこうと思います(せっかく作ったのだし,私自身のローカルPCでは動いているので・・・).すみません.
と,これではあんまりなので,Shakespeare の異綴字を比較した結果を披露しておきます.「#1720. Shakespeare の綴り方」 ([2014-01-11-1]) で挙げた25種類の異綴字 Shakespeare, Schaksp, Shackespeare, Shackespere, Shackspeare, Shackspere, Shagspere, Shakespe, Shakespear, Shake-speare, Shakespere, Shakespheare, Shakp, Shakspe?, Shakspear, Shakspeare, Shak-speare, Shaksper, Shakspere, Shaxberd, Shaxpeare, Shaxper, Shaxpere, Shaxspere, Shexpere を入力して,ソートさせると,次のような出力が得られた.
Similarity | Spellings |
---|---|
1.0000 | Shakespeare |
0.9565 | Shackespeare, Shake-speare, Shakespheare |
0.9524 | Shakespear, Shakespere, Shakspeare |
0.9091 | Shackespere, Shackspeare, Shak-speare |
0.9000 | Shakspear, Shakspere |
0.8571 | Shackspere |
0.8421 | Shakespe, Shaksper |
0.8000 | Shagspere, Shaxpeare, Shaxspere |
0.7368 | Shaxpere, Shexpere |
0.7000 | Shakspe? |
0.6667 | Schaksp, Shaxper |
0.6250 | Shakp |
0.5263 | Shaxberd |
昨日の記事「#3397. 後期中英語期の through のワースト綴字」 ([2018-08-15-1]) に引き続き,今回は such の異綴字について.「#2520. 後期中英語の134種類の "such" の異綴字」 ([2016-03-21-1]),「#2521. 初期中英語の113種類の "such" の異綴字」 ([2016-03-22-1]) で見たように,私の調べた限り,中英語全体では such を表記する異綴字が212種類ほど確認される.そのうちの208種について,昨日と同じ方法で現代標準綴字の such にどれだけ類似しているかを計算し,似ている順に列挙してみた.
Similarity | Spellings |
---|---|
1.0000 | such |
0.8889 | sƿuch, shuch, succh, suche, sucht, suech, suich, sulch, suuch, suych, swuch |
0.8571 | suc |
0.8000 | sƿucch, sƿuche, schuch, scuche, shuche, souche, sucche, suchee, suchet, suchte, sueche, suiche, suilch, sulche, sutche, suuche, suuech, suyche, swuche, swulch |
0.7500 | sƿuc, scht, sech, shuc, sich, soch, sueh, suhc, suhe, suic, sulc, suth, suyc, swch, swuh, sych |
0.7273 | sƿucche, sƿuchne, sƿulche, schuche, suecche, suueche, suweche |
0.6667 | hsƿucche, sƿche, sƿich, sƿucches, sƿulc, schch, schuc, schut, seche, shich, shoch, shych, siche, soche, suicchne, suilc, svche, svich, swche, swech, swich, swlch, swulchen, swych, syche, zuich, zuych |
0.6154 | swulchere |
0.6000 | asoche, sƿiche, sƿilch, sƿlche, sƿuilc, sƿulce, schech, schute, scoche, scwche, secche, shiche, shoche, sowche, soyche, sqwych, suilce, sviche, sweche, swhych, swiche, swlche, swyche, zuiche, zuyche |
0.5714 | sec, sic, sug, swc, syc |
0.5455 | aswyche, sƿicche, sƿichne, sƿilche, sƿi~lch, scheche, schiche, schilke, schoche, sewyche, sqwyche, sswiche, swecche, swhiche, swhyche, swichee, swyeche, zuichen |
0.5333 | sucheȝ |
0.5000 | sƿic, sɩͨh, scli, secc, sick, silc, slic, solchere, sulk, swic, swlc, swlchere, swulcere |
0.4444 | sƿilc, sclik, sclyk, suilk, sulke, suylk, swics, swilc |
0.4000 | sƿillc, sclike, sclyke, squike, squilk, squylk, suilke, suwilk, suylke, swilce, swlcne, swulke, swulne |
0.3636 | sƿilcne, suilkin, swhilke |
0.3333 | swisɩͨhe |
0.2857 | sik, sli, slk, sly, syk |
0.2500 | selk, sike, silk, slik, slyk, swil, swlk, swyk, swyl, syge, syke, sylk |
0.2222 | sƿilk, selke, sliik, slike, slilk, slyke, swelk, swilk, swlke, swyke, swylk, sylke |
0.2000 | slieke, slkyke, swilke, swylke, swylle |
0.1818 | sƿillke, swilkee, swilkes |
「#53. 後期中英語期の through の綴りは515通り」 ([2009-06-20-1]),「#54. through 異綴りベスト10(ワースト10?)」 ([2009-06-21-1]) で紹介したように,後期中英語の through の綴字は,著しくヴァリエーションが豊富である.そこではのべ515通りの綴字を集めたが,ハイフン(語の一部であるもの),イタリック体(省略符号などを展開したもの),上付き文字,大文字小文字の違いの有無を無視すれば,444通りとなる.この444通りの綴字について,現代の標準綴字 through にどれだけ近似しているかを,String::Similarity という Perl のモジュールを用いて計算させてみた.類似性の程度を求めるアルゴリズムは,文字変換の工程数 (Levenshtein edit distance) に基づくものである.完全に一致していれば 1.0000 の値を取り,まったく異なると 0.0000 を示す.では,「似ている」順に列挙しよう.
Similarity | Spellings |
---|---|
1.0000 | through |
0.9333 | thorough, throughe, throught |
0.9231 | throgh, throug, throuh, thrugh, trough |
0.8750 | thoroughe, thorought |
0.8571 | thorogh, thorouh, thorugh, thourgh, throȝgh, throghe, throght, throwgh, thrughe, thrught |
0.8333 | thogh, throu, thrug, thruh, trogh, trugh |
0.8000 | thoroghe, thoroght, thorouȝh, thorowgh, thorughe, thorught, thourghe, thourght, throȝghe, throghet, throghte, throighe, throuȝht, throuche, throwght, thrughte |
0.7692 | þrough, thorgh, thorou, thoruh, thourh, throȝh, throch, throuȝ, throue, throwg, throwh, thruch, thruth, thrwgh, thrygh, thurgh, thwrgh, troght, trowgh, trughe, trught |
0.7500 | thorowghe, thorowght |
0.7273 | thro |
0.7143 | þorough, þrought, thorȝoh, thorghe, thorght, thorghw, thorgth, thorohe, thorouȝ, thorowg, thorowh, thorrou, thoruȝh, thoruth, thorwgh, thourhe, thourth, thourwg, thowrgh, throȝhe, throcht, throuȝe, throwth, thruȝhe, thrwght, thurgeh, thurghe, thurght, thurgth, thurhgh, trowght, yorough |
0.6667 | þorought, þough, þrogh, þroughte, þrouh, þrugh, thorg, thorghwe, thorh, thoro, thorowth, thorowut, thoru, thour, throȝ, throw, thruȝ, thrue, thuht, thurg, thurghte, thurh, thuro, thuru, torgh, yrogh, yrugh |
0.6154 | þorogh, þorouh, þorugh, þourgh, þroȝgh, þroghe, þrouȝh, þrouhe, þrouht, þrowgh, þurugh, thorȝh, thorch, thoroo, thorow, thorth, thoruȝ, thorue, thorur, thorwh, thourȝ, thoure, thourr, thourw, thowur, throȝe, throȝt, throve, throwȝ, throwe, throwr, thruȝe, thrvoo, thurȝh, thurch, thurge, thurhe, thurow, thurth, torghe, trghug, yhurgh, yorugh, yourgh |
0.5714 | þoroghe, þorouȝh, þorowgh, þorught, þourght, þrouȝth, þrowghe, þurughe, thorowȝ, thorowe, thorrow, thorthe, thourȝe, thourow, thowrow, thrawth, throwȝe, thurȝhg, thurȝth, thurgwe, thurhge, thurowe, thurthe, yhurght, yourghe |
0.5455 | þrou, þruh, thow, thrw, thur, trow, yrou |
0.5333 | þorowghe, þorrughe, thorowȝt |
0.5000 | þorgh, þorou, þorug, þoruh, þourg, þourh, þroȝh, þroth, þrouȝ, þrowh, þurgh, þwrgh, ȝorgh, thorȝ, thorv, thorw, thowe, thowr, threw, thrwe, thurȝ, thurv, thurw, thwrw, trowe, yhorh, yhoru, yhrow, yorgh, yorou, yoruh, yourh, yurgh |
0.4615 | þhorow, þorghȝ, þorghe, þorght, þorguh, þorouȝ, þoroue, þorour, þorowh, þoruȝh, þorugȝ, þoruhg, þoruth, þorwgh, þourȝh, þourgȝ, þourth, þroȝth, þrouȝe, þrouȝt, þurgȝh, þurghȝ, þurghe, þurght, þuruch, dorwgh, dourȝh, drowgȝ, durghe, tþourȝ, thorȝe, thorȝt, thorȝw, thorew, thorwȝ, thorwe, thurȝe, thurȝt, thurew, thurwe, yorghe, yoroue, yorour, yourch, yurghe, yurght |
0.4286 | þorouȝe, þorouȝt, þorouwȝ, þorowth, þrouȝte, thorȝwe, thorewe, thorffe, thorwȝe, thowffe, trowffe |
0.4000 | þro |
0.3636 | ðoru, þorg, þorh, þoro, þoru, þouȝ, þour, þroȝ, þrow, þruȝ, þurg, þurh, þuro, þuru, ȝoru, ȝour, torw, twrw, yorh, yoro, yoru, your, yrow, yruȝ, yurg, yurh, yuru |
0.3333 | þerow, þerue, þhurȝ, þorȝh, þorch, þoreu, þorgȝ, þoroȝ, þorow, þorth, þoruþ, þoruȝ, þorue, þorwh, þouȝt, þourþ, þourȝ, þourt, þourw, þroȝe, þroȝt, þrowþ, þrowȝ, þrowe, þruȝe, þurȝg, þurȝh, þurch, þureh, þurhg, þurht, þurow, þurru, þurth, þuruȝ, þurut, ȝoruȝ, ȝurch, doruȝ, torwe, yoȝou, yorch, yorow, yoruȝ, yourȝ, yourw, yurch, yurhg, yurht, yurth, yurwh |
0.3077 | þorȝhȝ, þorowþ, þorowe, þorrow, þoruȝe, þoruȝt, þorwgȝ, þorwhe, þorwth, þourȝe, þourȝt, þourȝw, þourow, þouruȝ, þourwe, þrorow, þrowȝe, þurȝhg, þurȝth, þurthe, ȝoruȝt, yerowe, yorowe, yurowe, yurthe |
0.2857 | þrorowe |
0.2000 | þor, þur |
0.1818 | þarȝ, þorþ, þorȝ, þore, þorv, þorw, þowr, þurþ, þurȝ, þurf, þurw, ȝorw, ȝowr, dorw, yorȝ, yora, yorw, yowr, yurȝ, yurw |
0.1667 | þerew, þorȝe, þorȝt, þoreȝ, þorew, þorwȝ, þorwe, þowre, þurȝe, þurȝt, þureȝ, þurwȝ, þurwe, dorwe, durwe, yorwe, yowrw, yurȝe |
0.1538 | þorewȝ, þorewe, þorwȝe, þorwtȝ |
Powered by WinChalow1.0rc4 based on chalow