Question Answering System Research And Application In The Field Of Carbon Trading

Question answering system has been a popular topic in the past years of artificial intelligence renaissance,representing an advanced technology of natural language processing.Differ from traditional search engines,question answering approaches allow users to express arbitrarily complex information needs in natural language.The non-compliant of questions is the main challenge for the question answering system,therefore question understanding is a key step in a natural language question answering system and is crucial to the success of the question answering system.In this paper,a question understanding model is proposed based on deep learning,and the practicality of question understanding model is verified by constructing a question answering system for carbon trading.The specific work and achievements are as follows:(1)Question understanding involves intent detection and constraint extraction.A BLSTM-CNN-CRF architecture for joint modeling of intent detection and constraint extraction is proposed based on the advantages of deep learning.Each word-level label is greedily determined by the hidden representation from the Bi-directional Long Short-Term Memory(BLSTM)network in each step.The semantic features of the whole utterance are automatically extracted through Convolutional Neural Network(CNN).Finally,the output vectors of them are fed to the Conditional Random Fields(CRF)layer to jointly decode the best label sequence.The joint model achieves competitive performance on the benchmark Airline Travel Information System(ATIS)task without any artificial features.(2)The construction of the frequently asked questions(FAQ)question answering system depends on the calculation of the similarity of questions.Therefore,a similarity calculation method based on question understanding is proposed.Intent similarity and constraint similarity are calculated by word embedding model.The final question similarity is obtained from the combination of the two similarities.Experiment results show that the similarity calculation method based on question understanding can effectively calculate the similarity between questions.(3)Through question understanding and semantic similarity calculation,the FAQ question answering system for carbon trading is implemented.Experiment results show that the system can return specific answers based on the questions asked and provide similar questions.