I am making a multi-class classifier based on the text of each employee to predict top k skills for that particular person. Skills are predefined and has to be chosen from that only. And since, the dataset is huge, the model should not have very large computation time. Also, I might not have a training dataset or it is not very accurate. So, Is it possible to build a model without training data , or any model exists that i can use for this task? Any help would be appreciated!