想要寫作28+可以了解下ETS 的e-rater 是如果操作的

不是什麽(me) 內(nei) 部機密，都是ETS自己的官網公布的東(dong) 西。

明白了這些機理以後，相信不管是口語還是寫(xie) 作，拿個(ge) 高分，你應該就會(hui) 比較有信心了。

當然就像課上反複強調了，了解怎麽(me) 打分，隻是指明了一個(ge) 明確的方向，當有了方向以後，剩下的就是練習(xi) 了。

...

首先說下TOEFL 寫(xie) 作的閱卷操作，內(nei) 容來自於(yu) ETS研究員 Catherine S. Trapani 的一篇關(guan) 於(yu) e-rater 的報告。

對於(yu) 不了解e-rater 的首先和大家簡單介紹下這個(ge) 軟件

# 關(guan) 於(yu) e-rater

e-rater 是 ETS 研發的一個(ge) 自動打分軟件（automated scoring model）。在 TOEFL 的寫(xie) 作，GRE 的寫(xie) 作和GMAT 的寫(xie) 作部分裏麵都有應用。

目前TOEFL 的寫(xie) 作部分的打分機製是，一位 human-rater 和 e-rater 打分，最終成績是兩(liang) 個(ge) 加起來的平均分。當然如果兩(liang) 個(ge) 的分值相差過大，會(hui) 有第二個(ge) human-rater 來打分。

E-rater’s use was investigated as a contributory score. Under the contributory score model, e-rater score was checked for agreement with the first human score within an empirically established range, beyond which a second human score was required. The average of the human and e-rater scores became the final score for the essay, unless a second human rating was desired.

另外要注意的是e-rater 的閱卷是模仿人的打分方式，所以通過e-rater 的打分機製，可以很好的知道，整個(ge) 寫(xie) 作部分到底是怎麽(me) 打分的。

# e-rater 打分機製

想要寫(xie) 作28+，可以了解下ETS 的e-rater 是如果操作的

這張圖非常的關(guan) 鍵，可以看出e-rater 的打分大體(ti) 分成幾個(ge) 角度

1. organization

2.development

3. positive features

4. grammar

5. usage

6. mechanics

7. style

...

當然這麽(me) 多內(nei) 容，第一個(ge) 問題：

這些名詞到底是什麽(me) 意思？？？

我們(men) 用paper 中的一段話來簡單地介紹下上麵幾個(ge) 概念。

Grammar, usage, mechanics, and style together identify over 30 error types, including errors in subject-verb agreement, homophones errors, misspelling, and overuse of vocabulary. These error types are summarized for each feature as proportions of error rates relative to the essay length. Organization and development features are based on automatically identifying sentences in an essay as they correspond to essay-discourse categories: introductory material (background), thesis, main ideas, supporting ideas, and conclusion. For the organization feature, e-rater identifies the number of elements present for each category of discourse in an essay. For the development feature, e-rater computes the average length for all the discourse elements (in words) in an essay. Lexical complexity of the essay is represented by two features. The first is computed through a word frequency index used to obtain a measure of vocabulary level. The second feature computes average word length across all words in the essay and uses this as an index of sophistication of word usage. A new feature indicative of correct use of collocation and preposition use in the essay was the first feature to be included in e-rater version 10.1 to support further development of measures of positive attributes of writing style and ability (Ramineni, Davey, & Weng, 2010).