MIT OpenCourseWare


» 進階搜尋
 課程首頁
 教學大綱
 教學時程
 相關閱讀資料
 作業

作業


本頁翻譯進度

燈號說明

審定:無
翻譯:林恭印(簡介並寄信)、賀陽劍(簡介並寄信)、李國玉(簡介並寄信)
編輯:侯嘉玨(簡介並寄信)

這門課程的句子處理部分有三個作業和一個問題範例。


語言處理:句子理解

作業1:

提交三份3頁的報告,簡要評論在某節課上提到的一篇或者幾篇論文。其中一篇報告需要在討論這篇論文的那一天在課堂上做20-30分鐘的口頭報告。

除了列在這裡的論文外,也可以就同樣主題的其他論文做口頭報告。請最遲在第三次上課前到我這裡確認各位所選的題目。

作業2:

將要發給大家的問題集占10%,最遲在第三次上課前完成。

作業3:

期末計畫/報告定於這門課程的最後一天繳交。如果提交的是一篇報告,頁數應該有10-20頁,報告可以針對句子處理這一領域內的任何主題。

論文可以由新提議、分析和數據組成,也可以是針對某一具體領域的評論性文章,包括評估和評論。

問題:句子處理實驗的設計

針對下面兩個問題,描述自定步調(或者眼球追蹤)閱讀實驗中項目的組成。描述項目中需要控制的變量類型。對每個實驗寫兩個範例項目。這些項目不需要控制詞頻差異,但是你需要說出項目中的哪些字詞需要與實驗中的實際詞頻相匹配。針對此設計好的實驗,描述比較的關鍵區域。

1. 設計一個實驗,檢驗插入性整合的複雜度除了影響插入語段結構的複雜度外,是否也影響了整合的難度。理想的對比是主詞摘錄關係子句和受詞摘錄關係子句之間的對比,如句子(1)和(2):

(1) 主詞摘錄關係子句:The reporter who attacked the senator admitted the error. (那個攻擊參議員的記者承認那個錯誤)
(2) 受詞摘錄關係子句:The reporter who the senator attacked admitted the error. (參議員攻擊的那個記者承認那個錯誤)

在處理「admitted」這一點上,這個動詞與主語「the reporter」是整合在一起的。在這兩個句子中,因名詞片語主詞而引入的語段是相同的,但是受詞摘錄關係子句中的這個動詞仍然比較複雜(讀起來會慢一些)。這表明插入區域的整合複雜度(句2比句1複雜)影響了交叉整合的複雜度。

以不同的結構,設計另外一個實驗來檢驗這一假設。

2. 有些研究者(如Fodor & Inoue,1994)認為,句(3)(其中含可消解歧義的句法線索)的重析複雜度小於句(4)的重析複雜度(這個句子中,重析的線索來自情境的不可能:桌子不可能裝進嘴裡)。

(3) I put the candy on the table into my mouth.(我把在桌上的糖果放進我嘴巴中)
(4) I put the candy on the table in my mouth.(我把糖果放入在我嘴巴中的桌子上)

請為測試這一觀點的實驗設計項目。

3. 從屬位置理論假設計算資源的使用由兩部分組成:儲存和整合。設計一個實驗,用以測試這兩種資源是附加的還是互動的。(如果它們是互相作用的,兩者使用的工作記憶資源或許會有重疊。)




Here are three of the assignments and an example problem in sentence processing for this course.


Language Processing: Sentence Comprehension

Assignment 1:

Three 3-page papers, briefly reviewing one or more papers to be covered in one of the classes. One of these papers will be presented orally for 20-30 minutes in class on the day that this paper is to be discussed.

Papers other than the ones listed here (but on the same topic) can be considered for presentation. Please get my approval on your choice. Due in Lecture 3.

Assignment 2:

10% for a problem set. Problem set to be handed out. Due in Lecture 3.

Assignment 3:

Final project/paper, due on the final day of classes. If this is a paper, it should be roughly 10-20 pages. This paper can be on any topic in sentence processing.

This paper may consist of novel proposals, analysis and data, or alternatively it may consist of a review article in a specified area, with evaluation and commentary.

Problem: Designing a Sentence Processing Experiment

Describe the form of the items to be used in a self-paced reading experiment (or eye-tracking) to address each of the following two questions. Describe the kinds of things that need to be controlled in the items. Write two example items for each experiment. These items do not need to be controlled for lexical frequency differences, but you need to say which words in the items need to be matched for lexical frequency in a real experiment. Describe the critical regions of comparison for the proposed experiment.

1. Design items for an experiment which would test whether the complexity of an intervening integration affects the difficulty of making an integration, in addition to the complexity of the intervening discourse structure. One comparison that has the desired properties is the comparison between subject and object-extracted relative clauses, as in (1) and (2):

(1) Subject-extracted RC: The reporter who attacked the senator admitted the error.
(2) Object-extracted RC: The reporter who the senator attacked admitted the error.

At the point of processing "admitted", this verb is integrated with its subject "the reporter". The discourse objects that have been introduced since this subject NP are the same in the two sentence structures, but the object-extracted RC version is still more complex (slower to read) at this verb. This suggests that the integration complexity of the intervening region (which is more complex in (2) than in (1)) affects the complexity of crossing integrations.

Design items for another experiment testing this hypothesis, on a different construction.

2. It has been claimed by some researchers (e.g., Fodor & Inoue, 1994) that the reanalysis complexity of (3) (in which there is a syntactic cue to ambiguity resolution) is less than the reanalysis complexity of (4) (in which the cue to reanalysis comes from the implausibility of the situation: a table can't be inside a mouth):

(3) I put the candy on the table into my mouth.
(4) I put the candy on the table in my mouth.

Design items for an experiment testing this claim.

3. The dependency locality theory hypothesizes that there are two components to computational resource use: storage and integration. Design an experiment that tests whether the two kinds of resources are additive or interact. (If they interact, there may be an overlap in the working memory resources used for each.)




 
MIT Home
Massachusetts Institute of Technology Terms of Use Privacy