The Design Of The Application To Engineering Data Based On Intelligent Assistant

With the development of artificial intelligence,many companies launch a series of intelligent assistants.These technologies and products greatly influence people's life and enable humans and computers to communicate directly.That people interact with the machine by natural language may have potential becoming a mainstream way of Human-Computer Interaction.Based on this situation,there exists huge prospects for intelligent assistants in the field of engineering application.In the workshop,commanding intelligent assistants in natural language can replace the keyboard and touch screen as an alternative method to help searching some lightweight engineering data.This project is about of applying Amazon's smart assistant Alexa in processing engineering data,with the aim of creating an Alexa skill as the application to process engineering data.The user can use the voice to command intelligent assistant to perform query,add and modify operations on the engineering data in the database,and can perform the functions of back and help.Amazon has provided developers amounts of platforms and products for developing Alexa skills.Amazon Developer Platform can be used to help creating the interaction model of the skill Lambda function can be used to provide computing service and generate the response,and Dynamo DB can be used to save all engineering data.This thesis uses the above three tools to develop different parts of the skill and combine these parts to form the final skill.After testing,the designed Alexa skill in this thesis can achieve the required functions.In addition,there are some shortcomings in Alexa's interaction with the user,especially in terms of user's command intent recognition.The skill developer must prepare a complete set of examples of the command utterances.When the user's utterance conforms to the format of the command example,Alexa can recognize the intention of the user's utterance.This has caused many commands that do not conform to the format cannot be recognized by the intelligent assistant.Aiming at the insufficient ability to identify the intent of user commands for intelligent assistant with interacting with users,a command text intent recognition model based on the Bidirectional Recurrent Neural Network(Bi-RNN)of Gated Recurrent Unit(GRU)is proposed in this paper.Firstly,the command text is transformed into a word vector model with the word embedding model added as the input of the command text intent recognition model.Secondly,the semantic characteristics of the text are analyzed by using the recurrent neural network.Finally,the result is classified by the Softmax classifier to complete intent recognition.In order to verify the accuracy of the proposed model,this paper uses the command examples of Amazon Alexa skill and Maluuba as the data set for experimental testing.Through the comparative analysis with other intent classification and recognition models,the accuracy of the intent recognition and optimization model designed in this thesis is obviously improved.The experimental results show that the proposed model can achieve a recognition rate of up to 94.7%,which can effectively improve the ability of intelligent assistants to recognize user commands and improve the overall performance of traditional intelligent assistant products.