Lesson 12 Deep Learning on Natural Language
Natural language means our languages, which we use everyday. That includes sentences written in Japanese or English. We have a variety of examples AI utilizes natural language. Translation to the other language, chatbot, which automatically creates answers to human beings' questions, are the part of them. Natural language area is gathering attentions, as the performance has been dramatically improved by Deep Learning.
Now, we are explaining the process in natural language.
1 Learning Data Set Preparation
Collect data sets of translation source and target. If you focus on translation from Japanese to English, the source language is Japanese, and the target language is English. Chatbot will ask you to prepare Japanese for both the source and the target.
2 Learning Execution
Give teaching data set to the system, and adjust parameters in the system so that the output value is going to be closer, as image recognition processes. Make the system to calculate repeatedly, using a lot of data sets of input and output sentences. The system will gradually get more probable answers to any input sentences.
3 Result Evaluation
Confirm the result of the system by testing input senteces. The data set for this purpose is called Testing Data Set. You can see the accuracy of natural language process by learning.
