フィンランドで取った授業のまとめです。
先生(Ken Pennington)の母語はアメリカ英語だけど、語法の専門家らしくて、国や分野ごとの語法の違いも指摘したとても含蓄のある授業でした。

構成を考える。内容的には

• situation (現状)
• problem (問題点)
• solution (解決策)
• evaluation (評価)

となる。
構成としては、

• introduction(導入)
• methods (方法)
• results (結果)
• discussion (議論)

記述上の特徴としては、

• Introduction : 現在・引用(citation)・注釈 (commentary)
• Method : 過去・受け身
  
• Result : 過去
• Discussion : 現在・引用(citation)・譲歩(qualification)

現在形・現在完了が使われる例が多い。

• Recently, there has been growing interest on...
• The possibility of ... has generated wide interest on...
• Many investigators have recently turned to...

ネガティブなことを書く。
However, few data are ...
みたいに、However + few/littleを使うのが典型的。
プチネガティブ。

• concentrate on ...
• been limited to ...
• over/under-estimated ...
• been restricted to ...

だいぶネガティブ

• disregarded to ...
• failed/neglected to consider ...
• ignored ...
• misinterpreted ...
• overlooked ...

The aim of this paper is ...
現在形で書くのが普通になりつつある。 二つめの目的をつなげたければ、In additionとかSecondlyとかで続ける。

例えばこんな例。動詞を変えているのに注意。
The plan of this paper is as follows. Section II describes .... In Section III, a theoritical model is constructed which ... ... are then tested in Section IV. Finally, Section V offers some suggestions for ...

命令形はあまり使わない。でもletはよく使う。
他には、

• Suppose that...
• Consider that...
• Note that...
• Compare that...

は文法的にはOK。Notice...は誤り。
命令形は簡単に書き換えが出来る。

• Now compare the result in table 4 and 5.
• →The results in table 4 and 5 can now be compared
• →If we now compare the results of table 4 and 5, we can see...

副詞は原則動詞の近くに置く (be動詞の後ろ、普通の動詞の前)
to不定詞の場合は、to sharly riseみたいに挟むか、to rise sharplyって書くか意見が分かれる

• First, finally, in addition, nevertheless (にもかかわらず)は絶対文頭に置く
• moreover, thus, in particularは80%くらいの確率で文頭
• however, for instanceは挟むことも多い
• also, for example, therefore, thenは文頭より文中の方が普通
• Moreoverは、howeverのあとに使う。"However 問題1. Moreover 問題2"みたいに

Howeverの位置は肯定文・否定文にかかわらず文頭が多いようです。

• 文 + 文はピリオド(.)で切るのが普通だけど、関連している文はセミコロン(;)でつなぐ。
  典型的な例はこんな感じ。

  ABC will provide an estimate of this cost increase; traditional costing will not. 
  

  ; however, とか、; for example, みたいな使い方もOK。
• 比較的長いものを並列させる時に、;で区切る。

   Transactions can be scored discriminatingly on three dimensions:
  (1) asset specificity;  (2) uncertainty (including complexity, 
  which is similar to uncertainty in its effects); and (3) frequency.
  

• リファレンスを並べる。カンマを著者と年号を並べるのに使ってるから、それより強いのが必要。

  Several authors mention the use of an ABC analysis as a basis for 
  performing a VCA (Guilding et al., 2000; Mecimore and Bell, 1995;  
  Shank and Govindarajan, 1992, 1993).
  

大体感覚的に分かってることだけど。
In additionとかthusとかの後にカンマを忘れないように。

• Kei, who lives in Finland, wrote...

  情報の追加 (非限定用法)

• However, ...

  文をつなぐ言葉の後 : however, furthurmore, in addition, as a result, in contrust...

• In figure 10, ..

  in, on, toとか、前置詞が前に出るケース

• Wherever you go, 

  従属接続詞。

• Browsing through the journal 'Science', 

  ､ 現在分詞の文。

• ... , resulting ...

  現在分詞が後に来るとき。動詞を並列するとき( and result in ...) はカンマは来ない。

• , and / but/ so

  等位接続詞の後

• Nokia, one of the biggest company in Finland, is ...

  省略できる説明の句

• apple, orange, and banana

  いろいろならべるとき。and の前に,を入れないのがイギリス流、入れるのがアメリカ流だけど、入れたほうが意味がはっきりする

• Sushi is an exotic, expensive cuisine for Finnish people

  形容詞をand無しで並べる

• × : 短縮 (can't, don't)
• × : and, butから始まる文
  →however, furthermore, in addition
• × : 命令形・疑問文
  →間接疑問 : We now need to consider how costs may be lower..
• × : I, youから始まる文
  →受動態
• × : stuff, things, bunch, a lot of, kind of
• × This kind of ...
  → such ...
• × : not ... any/much/many,
  →no, little, few
• × : A, B, C, etc.
  →devices, such as robots, and CD players
  まとめの言葉, such as/ including A, B and C
• 主語を工夫して文を完結に。
  It is difficult to make a decision about... → Choosing ... is difficult

bring up	もたらす	cause
look into	検証する	investigate
figure out, find out	明らかにする	determine
come up with	開発する、思いつく	develop, devise, invent
make up	構成する	constitute
get rid of	取り除く	eliminate
go up to	到達する	reach
keep up	維持する	maintain
go down	減少する	decrease, diminish, drop (reduction)
go up	増加する	increase, augment
look over	もう一度見る	review
run into	遭遇する	encounter, face
bring up	提起する	raise, present
look at	見る、精査する	examine

• A cause B は、Bが悪い事の時に使う
• A yield B はあまり乗り気でない感じ。既存研究の紹介とかで使う
• implement 実行する・施行する。プログラムだと「実装」だけど。
• assumtion は思いこみみたいなニュアンスがあって、間違っていることが示唆されるかも
• approach はwayのフォーマルな言い方として使える
• decline は「減らす」だから、「減る」というcontextでは使えない。
  reduction は「減る」だから、「減らす」というcontextでは使えない。
  decreaseはどちらでも使える。
• A total of 467 sentence connectors was found...
  全部で、というかっこいい言い方。
• ハイフンを用いて、係り受けをはっきりさせることができる。
  small car factory → small-car factory
• This, they, thatの使い分けはなかなか難しいけど、基本パターンは下。

  X is Y. This Y is ...
  

• 冠詞はやっぱり奥深い。
  A computer and printer is on sale for $1... コンピュータとプリンタが(合計)1ドルで売られている
  A computer and a printer are on sale for $1... コンピュータとプリンタが(それぞれ)1ドルで売られている

The grid is a computing environment consisting of a number of computers, typically PCs. SuchThosecomputers are spreaded distributed in many places locations, and connected with to the internet. Comparing Compare to a supercomputer, this system is built quite cheaper offers cheaper price. While the flexibilty of this environment in terms of operations and scale is quite high, their the failure rate of this system is unnegligibly sufficiently high. Thus, a special program with tolerance for failures is needed. Thus, a grid environment requires a special program that tolerates failures.

As a consequence of the growth of the internet, websites has become one of the most important commercial media. Generally It is known that good design, especially color design, attracts customers, but it designing such sites is not an easy task for most people.

We propose an automatic color advising system for websites, using a commonly used method called "machine learning". With this system, a user can create This system allows users to create an infinite number of designs from based on existing well-designed websites.

Now we proceed to the detailed process. The advisory system consists of three-steps process. First, a user gathers websites with good color patterns; those thses websites are used as material data for the system. Second, those the data are processed with a machine learning system, which extracts general criterion about the color design from them the data; now as a result, the system can tell descriminate good color design and bad one from less effective designs. Third, the color designs are randomly generated and evaluated by the machine learning system. If a design is determined to be satisfactory, that candidate is output. the system outputs that candidate design.

Table 16 illustrates a comparison of between the Regular and the Makeup Exams. As obviously can be seen, the average score of in the Regular exam is significantly higher than that of in the Makeup exam: 86 in the former comparing compared to 72 in the latter. However, this difference was not only conducted by can be attributed to inadequate the greater difficulty in of the Makeup exam. Now we consider the other reasons. Other reasons may have also contributed to the difference in scores.

First, the room condition was worse in the Makeup exam. The temperature was 28 degrees, comparing compared to 20 degrees in the Regular exam.

Second, we cannot overlook the possibility of cheating in the Regular exam; , since there were 125 students, which is 5 times more than that of the Makeup Exam, for one proctor. This ratio of proctor and to examinees werewas not ideal leastways.

First, the program definitions of a fraction object is written the program writes the definitions of a fraction object. On the writing of methods When the method is written, a care needs to be paid not to access to avoid accessing to remote data without remote method invocation (RMI). The system generates the complete object from the fraction definition. Then the main routine is written with using the object. The second step involves using the object in the main routine. When a method of the object is called remotely, a special grammer is used. Finally, all the codes are processed by our preprocessor and the parallel program is obtained to obtain the parallel program

In many data-intensive applications, each node can start processing as soon as it has received required data. Thus, each node is desired to receive data in the largest broadest-possible bandwidth. Broadcast algorithms are usually evaluated by the longest completion time among all nodes, but this criterion only focuses on the slowest node. Instead, we believe that aggregate bandwidth, the cumulative sum of data that each node receives receiving by all nodes in a unit time, better describes the performance of a long-message broadcast. Under this criterion, we call say a broadcast is stable when the aggregate bandwidth of some nodes are not diminished by adding other nodes. We propose a stable broadcast algorithm that uses multiple partial pipelines. Under the assumption of a tree symmetric network, it is proved that each node can receive as much amount of data as in the exclusive direct transfer from the source. We proved that, under the assumption of a tree symmetric network, our broadcast algorithm delivers the same amount of data to each node as when an exclusive data transfer is performed to it. As a result, the aggregate bandwidth is maximized. Our simulation has shown that our algorithm achieves the best performance and is stable for adding nodes with narrow bandwidth. In a high large bandwidth variance environment, it performed twice aggregate bandwidth comparing to a single depth-first pipeline. our schema yielded twice the aggregate bandwidth in comparison to a single depth-first pipeline. We have also performed a experiment on the a real environment to assure demonstrate its practicality.