Dataset for word generation (Algorithm 1 validation dataset)

The dataset contains two sets of data: Chinese and Czech. The open-source Chinese dataset consists of 26 valid subject pairs (52 individuals in total). Each subject's data file includes the following:Experimental record information, such as the type of task and trial details;The prior context relied upon during the word prediction process;Words predicted by the GPT-2 model during the experiment, including the predicted word frequencies, parts of speech, word vectors, and other relevant information;The ambiguous words actually spoken by the subjects during the experiment, along with their frequencies, parts of speech, and word vector-related information;The cognitive model residuals of the predicted words, used to fit the Bayesian word prediction algorithm.The second dataset is in Czech and includes 20 valid subject pairs (40 individuals in total). The experimental tasks and data preprocessing methods, as well as the content and format, are consistent with those of the Chinese dataset, so they are not elaborated here.

本数据集包含两组数据，分别为中文数据集与捷克语数据集。其中中文开源数据集包含26组有效被试对（总计52名被试）。每名被试的数据文件均包含以下内容：实验记录信息，例如任务类型与试次细节；词汇预测过程中所依托的先验上下文；实验中GPT-2模型（GPT-2）生成的预测词汇，涵盖预测词频、词性、词向量及其他相关信息；实验中被试实际说出的歧义词汇，及其词频、词性、词向量相关信息；用于拟合贝叶斯词汇预测算法的预测词汇认知模型残差。第二组数据集为捷克语数据集，包含20组有效被试对（总计40名被试）。其实验任务、数据预处理方法、内容与格式均与中文数据集一致，故不再赘述。