AUTOMATIC LINKAGE OF VITAL RECORDS
AUTOMATIC LINKAGE OF VITAL RECORDS
复制标题
DOI：
10.1126/science.130.3381.954
复制
发表时间：
1959-01-01
期刊：
影响因子：
56.9
通讯作者：
JAMES, AP
中科院分区：
文献类型：
--
作者：
NEWCOMBE, HB;KENNEDY, JM;JAMES, AP
主要关心的是发展一种完全自动化和不需要零碎操作的程序，因为零碎操作以后可能会限制这种方法的有用性。这一目的得以实现，主要是因为计算机的使用使得将每个出生记录依次与婚姻档案中相应部分的所有婚姻记录进行比较成为可能。由于有时对一组婚姻进行多次扫描，很明显，这种操作无法用传统的卡片处理设备进行。因此，如果没有计算机，就需要通过印刷清单进行视觉搜索才能实现某些联系。为了减少计算机必须与出生记录进行比较的婚姻记录的数量，决定同时使用丈夫的姓氏和妻子的婚前姓氏，这两个姓氏同时出现在结婚卡和出生卡上。姓氏首先被简化为语音代码，由姓氏的第一个字母和三个数字组成，被称为罗素Soundex代码(8)，用于编码操作的计算机。这些代码有两个目的：它们被设计成与许多常见的拼写变化保持不变，因此在目前的应用中，它们被期望将可链接的记录聚集在一起，如果按照严格的字母顺序排列，这些记录将会被广泛地分开。编码还简化了Datatron计算机的后续使用，Datatron计算机本质上是一种数学工具，处理数字比处理字母更容易。两个姓氏在多大程度上比一个姓氏更有效地识别一个家庭群体，这一点可能尚未得到普遍认可。因此，在婚姻档案中遇到的各种新娘的婚前姓名中，有一半以上（以编码形式）重复出现，频率从64到1024 / 105不等。与此相反，近80%的姓氏对（以编码形式）是唯一的；也就是说，它们在我们的文件中以特定的组合只出现一次，并且极少数频率超过每10次4次（见图1）。这
Methods Of primary interest was the develop-ment of a procedure which would be fully automatic and free from piecemeal operations which might later limit the usefulness of the approach. This aim was achieved, chiefly because the use of a computer made it possible to compare each birth record in turn with all of the marriage records in appropriate sec-tions of the marriage file. Since groups of marriages were sometimes scanned a number of times, it is apparent that this operation could not have been carried out withconventional card-handling equipment. Thus, without the computer, a visual search through printed lists would have been required to achieve some of the linkages. To reduce the number of marriage records with which the computer must compare a birth record, it was decided to make use of both thehusband's surname and the wife's maiden name, these being present on both the marriage and the birth cards. The surnames were first reduced to phonetic codes, consisting in each case of the first letter of the name followed by three numeric digits and known as the Russell Soundex Code (8), the computer being used for the coding operation. The codes served two purposes: They were designed to remain unchanged with many of the common spelling variations and in the present application were thus expected to bring together linkable records which would have been widely separated if arranged in a strictly alphabetic sequence. The coding also simplified the subsequent use of the Datatron computer, which is essentially a mathematical instrument and works more readily with numbers than it does with letters. The extent to which twosurnames are more efficient than one for identify-ing a family group has probably not been generallyrecognized. Thus, of the various brides' maiden names encoun-tered in the marriage file, more than half recurred (in their coded forms) with frequencies in the range from 64 up to 1024 per 105. In contrast to this, nearly 80 percent of the pairs of family names (in their coded forms) were unique; that is, they occurred only once in our file in that particular combina-tion, and extremely few had frequencies exceeding 4 per 10 (see Fig. 1). This