Legal Judgement Prediction and Analysis

Project Title

Legal Judgement Prediction and Analysis

Project Area of Specialization Artificial IntelligenceProject Summary

Nowadays, when so many courts adhere to the directive to promote accessibility and re-use of public sector information and publish cases online even in Pakistan where we can find all the cases online on their website, we can easily analyze the data to extract meaningful information from it and use it for different purposes. In this case we are using machine learning to perform quantitative analysis to predict the decision of the Courts. If we can predict the results adequately, we may subsequently analyze which words made the most impact on the decision and this identify the factors that are important for the judicial decisions.

when courts started publishing judgements, big data analysis i.e. large scale statistical analysis of case law and machine learning with in the legal domain became possible. by taking data from pakistani courts, as an example, we investigate how Natural Language Processing tools can be used to analyze text of the court proceedings in order the automatically predict judicial decisions and helping judges while giving verdicts on the cases. Our approach highlights the potential of self learning approaches on the basis of previous data, in a legal domain. however, predicting decisions for future cases based of the cases from the past negatively impacts performance. Our semantics will be the previous judgements. we will use them to work on predicting decision or a verdict of a new case on the basis of it's facts. our project will include automatic summerization of legal information and information extraction. categorization of legal resources on the basis of artical or what type of case it is. statistical analysis of cases. whether the decision is converging towards the previous similar cases or diverging. we will also extract the features which were most useful in decision making. Finally our system will predict judgement. 

Project Objectives

Our objective in this projeect will be to perform quantative analysis of the basis of words and phrase. we will have a text judgemnt document as input. we will analyze it and process it by different techniques to help us get the meaning or sense out of it. it will be done on the basis of words and phrases in it.

Our system will extract the features which are important in decision making. We will check the relevence and occurance of important words or features. it will tell us about the relationship between verdict and facts.

we will train these features on a machine learning algorithm to check the accuracy and relevance. it will be trained to check the relationships between different facts. 

based on these analysis and learning, our model will be able to predict the decision of the court. when a new case will be directed, it will be fed into the system. system will then analyze it's facts and find revelance to previous traning data. it will then predict the decision or verdict label accordingly. 

Project Implementation Method Technique

Our implementation consists of the following steps:

Structuring Data

For machine learining and NlP techniques data must be highly structured so our first step was to sturcture the semi structure data which was in Text format. Structured data means data in rows and cols where rows represent documents and col represents attributes. In order to structure the data we divided the whole case into few cols which are:

• Date
• Court Name
• Judge Name
• Petitioner
• Respondent
• Judgement
• Verdict

Text Preprocessing

Text preprocessing is the important step when you are working on text data. Text preprocessing is consist of few techniques which are:

Removed white spaces removed extra whites spaces from data set which will be helpful while learning process. Lowering of data

converted all the data into lower case format so our model should predict better results.

Removed Unkown Characters & Punctuation

 Punctuation affectsmeaning of data and Deleting punctuation reduces the ability of the follow-on semantic parsing functionality. so we removed all the unknown characters and punctuations from data.

Removal of stop words

In sentences, there were many extra words i.e to,is, an, and etc. They are called Stop words so we removed these words using NLTK STOP WORDS Library.

Tokenize

This breaks up the sentences into words based on a specified pattern using Regular Expressions aka RegEx.

 Lemmatizing

 maps common words into one base. Unlike stemming though, it always still returns a proper word that can be found in the dictionary.

Feature Extraction

After preprocessing the data we applied Tf-Idf vectorizing the on the column of our interest. Tf give frequency of words and Idf give words that are important but have low frequency in the document and using this technique we removed some common words and organized data into more structured way in excel format.

Classification

After all data preprocessing is done we move on to apply classification techniques on data. For this data, we chose two of classifications one is Support Vector Machine (SVM) and another is Artificial Neural Network (ANN). And from data given in above table we can see that ANN give us more accuracy than SVM.

Benefits of the Project

Our system will benifit people and our governmental organizations. 

Courts and judges

our system will help Pakistani courts to keep the track of the cases. it will give the analyzed results of all the judgements based on the facts. Pakistani courts have a system of decision making in which Judges refer to previous decisions manually and go for a decision. they have to look up for the articles manually in a huge set of data. our system will  provide all the information with a click. it will perform all the analytics and calculations at back end which will save a lot of time. these will be way more accurate. 

Lawyers

System will also help lawyers in a way thay when a client comes up with a new case for them to fight. they can look up for the possibilities and refer to articles. on these basis, it will help them with preparing to defend in courts and providing references. it will help them to decide how to fight case whether to take this case or not.

General Public

General public can access our system through web based or app based portal vis simple login. it will help them find all the basic information, which they do not know as a layman. It will provide them with information and help them fight their case.

Technical Details of Final Deliverable Final Deliverable

Final deliverable will be a web based tool consisting of following features:

• User friendly interface 
• judge can search previous cases 
• judge can verify his decision 
• statistical analysis of predicition 

User friendly interface 

frontend of the interface will be designed using latest frontend tool to make it more interactive and user friendly. 

Judge can search previous cases 

we will provide a log of previous cases in our final product if user wants to take a glimpse of past cases he/she can simply search in the log maintend.

judge can verify his decision 

The important and main feature of product is to predict the court decsions. it will be a very  useful tool for verfying the judgments given some facts about the case.

statistical analysis of predicition 

While verfying the judgment our model will give following details about the case:

• How much this judgment deviates from the orignal decsion.
• Points on which the decsion has been predicted.
• Accuracy of the predicted results.
• Charges and penalty applied on the convict and articles voilated by the convict. 

Final Deliverable of the Project Software SystemCore Industry Legal Other IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Industry, Innovation and InfrastructureRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	60000
PC with a GPU	Equipment	1	60000	60000