Research On Key Technologies Of Intelligent Legal Aid System Based On Embedded System

With the development of artificial intelligence technology and the basic completion of legal informatization,intelligent legal aid system has become an important demand at present.Legal intelligence is to enable the machine to read the texts in the legal field and give preliminary pre-judgment according to the analysis of the contents of the texts,so as to help practitioners in the legal field complete their work more efficiently,help people better understand relevant legal knowledge and improve people's legal awareness.Legal intelligence has always been a hot research field.Early scholars proposed to use artificial features combined with machine learning classification algorithm to realize legal prediction.With the development of deep learning,the application of deep learning in the field of legal intelligence has become more and more popular,and many legal aid systems have been developed based on deep learning algorithms.Most legal aid systems are developed on server side with excellent computing performance,with poor portability and limited usage scenarios.Therefore,it is of great significance to study the legal aid system based on embedded terminal to promote the development of artificial intelligence in the legal field.This paper studies and designs a legal aid system based on embedded terminal.For the design of the whole system,the main contents and innovations of this paper are as follows:1.Investigate relevant legal websites and public legal data sets,study the Scrapy framework,design high-performance web crawlers,grab data,construct corpus of legal consultation questions,download open source data sets to construct corpus of crime prediction,and prepare for the later establishment of legal aid models.2.Introduce the construction process of legal aid model in detail,including data preprocessing,Chinese word segmentation,delete stop word,feature extraction and classification algorithm research.For feature extraction,according to the size of the corpus,Tf-idf and Word2 Vec are used for text expression.To avoid the lack of expression of professional vocabulary,Tf-idf and Word2 Vec models are trained according to the constructed corpus.3.For the selection of classification algorithm,considering the computing performance of raspberry pi,the Random Forest algorithm and the designed Convolutional Neural Network are selected as the basic classification algorithm.In view of the traditional models are all artificial design parameters,this paper proposes to apply the Sine and Cosine Algorithm to the parameter optimization process of the above model,and the SCA?RF and SCA?CNN obtained are used as the main prediction tools.4.Based on distributed task queue framework Celery,Web framework Flask,etc,combined with Sine and Cosine Algorithm,and SSH communication as the basis,a distributed hyperparameter optimization system is implemented,which is used for model training and optimization,and greatly shortens the model parameter searching time.5.Transplant the model to the embedded platform raspberry pi 4b.Raspberry pi 4b through microphone and written voice recording program,can be used to collect speech signals for speakers,combined with i FLYTEK speech recognition interface,to achieve a legal aid system based on voice recognition.