by Helen Johnson | May 11, 2017 11:13 am
Will artificial intelligence be able to ask questions? Probably, Maluuba knows the answer. Maluuba is a startup from Canada. The company is focused on creating the software that will be able to talk and even think like human beings. From the very beginning, the company was oriented to AI and investigating natural language.
Current voice/conversational assistants like Siri or Cortana operate rather good. But they are unable to put questions. Maluuba strives to improve the capacities and abilities of such systems. In January 2017, Microsoft acquired Maluuba and together they were focused on machine learning and the procedure of question generation.
The team planned to combine reinforcement and supervised learning. According to reinforcement learning approach, the machines and software agents will automatically define an optimal behavior depending on a particular context. This model will ensure the enlargement of machines performance.
An accurate and grammatically proper speech got the highest priority. The conversational assistant will generate the questions within the source text. The procedure of evaluating the quality of the questions is going to be rather interesting. A machine learning model will give answers to the questions generated by the assistant. And the accuracy level of answers will define the quality of the question generation model.
The project team is sure that the ability to generate questions will lead to the enhancement of the algorithm of their answering. It is easier to answer as usually, a question requires one specific reply. And it is possible to get a certain information by putting different questions.
The question generation machine will make grounds for the development and accomplishment of other machine learning models. Also, it is going to be used in the sphere of education. Pupils and students will be able to reply the questions automatically generated by the machines.
Source URL: https://blog.qatestlab.com/2017/05/11/maluuba-question-generator/
Copyright ©2019 QATestLab Blog