|Programs Normalization for More Precise Software Measurements=程式正規化：為軟體被更精確的量度
| 許多軟體度量是以程式中的運算子，運算元，邏輯判斷條件，變數的定義使用對(definition-use pairs)或是程式中的行(LOC)來當作是量度各種軟體性質的單元，軟體量度通常是被用來預測軟體的品質或者是用來估計開所發軟體的效率如何的工具，目前已經有許多著名的軟體度量被提出，這些度量衡是用來量度軟體中的各種屬性，我原以為這些度量可以量出軟體中正確的屬性值，但我發現在使用這些度量的過程中，有一項程序被忽略，使得所量出得的軟體屬性值有許多問題，這項程序便是人們在使用這些度量時，往往沒將一些不需要被量度的元素與結構排除在量度之外，使得所量得的值並不正確，這種將不需要被量度的元素與結構排除在被量度的程式之外的觀念很簡單但是很重要，可是確往往被忽略在絕大部份的量度過程中，因此我們在文中提出將軟體正規化的方法，此方法是將軟體轉化為較具規則化的結構這種結構將使得這些軟體被量度所得的數值較為合理。
| Software science provides interested sets of metrics at the levels of source code by using the number of LOC, operators, operands, predict conditions and definition use pairs presented in the code. Software measurement is used to predict software quality and to evaluate the effectiveness of design. There have been a great deal of interests in defining appropriate ways to measure the properties of software. We thought that these measures should accurately represent those attributes they purport to quantify. However, we identify a number of problems with these software measurements that are not precise enough because they always measure the elements that need not to be included. To reduce coarse and redundant structures of the program before being measured is a simple but important concept. However, this concept has been ignored in most software measuring activities. So, we propose a program normalization strategy that is the process of making programs to be in a regular form or a standard form to which others can be approached. This program normalization strategy can let those metrics present what they really want to measure and avoid the non-precise measured results.