Identifier Splitters

The oracle, found in the file loyola-udelaware-identifier-splitting-oracle (ludiso) consists of 2663 identifiers. These identifiers were split by volunteers using a web interface. In all 8522 splitting judgements were collected for 2731 identifiers. The data set was curated based on the sum of confidence associated with a particular split. In order to provide a single correct split in the oracle, the split receiving the highest confidence sum was used as the correct split. Of the 2731 identifers, 68 do not appear in the oracle because of ties. (These identifiers can be found in the raw data.) Details on the construction of the oracle can be found in "An empirical study of identifier splitting technique". Attribute Information Each line of the oracle represents one of the 2663 remaining oracle identifiers and includes the following. the identifier's unique identification number, the original identifier as extracted from the source, the dominant language of the program that the id was extracted from, the program from which it was extracted, the identifier as shown to subjects where a '-' denotes a hard split, the number of splittings (always 1 in the oracle, but not in the raw data), the identifier as split by users, the number of confidences (1 - 5) and then each confidence.

本神谕集（oracle）存储于文件loyola-udelaware-identifier-splitting-oracle（简称ludiso）中，共包含2663个标识符（identifier）。这些标识符由志愿者通过网页界面完成拆分。本次研究共为2731个标识符收集到8522条拆分判定结果。本数据集基于各拆分方案对应的置信度总和经校验整理而成：为在神谕集中得到唯一的正确拆分方案，我们将置信度总和最高的拆分作为最终正确拆分。在全部2731个标识符中，有68个因存在置信度总和持平的情况未被纳入本神谕集，此类标识符可在原始数据中查询到。关于神谕集的构建细节可参见论文"An empirical study of identifier splitting technique"。
属性说明：神谕集中的每一行对应2663个有效标识符之一，包含以下字段：标识符的唯一标识编号、从源代码中提取的原始标识符字符串、该标识符所属程序的主导编程语言、标识符所提取自的程序名称、展示给受试者的标识符形式（其中"-"代表硬拆分点）、拆分次数（在神谕集中恒为1，原始数据中则不满足此条件）、用户实际拆分后的标识符形式、置信度的数量（取值范围为1-5）以及各条对应的置信度分值。