WorksheetsCertified Legal Analytics Pretest
Total questions: 100
Worksheet time: 50mins
What is the primary purpose of legal analytics?
Automating court rulings
Enhancing legal decisions through data-driven insights
Replacing lawyers entirely
Standardizing all global legal systems
Which is an example of legal analytics?
Writing new laws
Forecasting litigation outcomes with past case data
Drafting contracts manually
Verifying client identities
Legal analytics primarily shifts legal practice from:
Evidence-based to intuition-based
Intuition-based to evidence-based
Civil law to common law
Statutory law to constitutional law
Preprocessing of legal data involves:
Drafting pleadings
Cleaning, deduplicating, and standardizing datasets
Writing statutes
Creating legal arguments
A key ethical challenge in legal analytics is:
Overuse of case law
Bias in datasets and predictive models
Lawyers refusing to learn coding
Too much client data digitization
Classification in ML assigns:
Continuous numerical predictions
Items into predefined categories
Similar items into clusters
Features into reduced dimensions
Predicting damages in a civil case uses:
Classification
Regression
Clustering
Association
Grouping similar contracts without labels is:
Classification
Regression
Clustering
Dimensionality reduction
The 'black box' problem refers to:
Opaque decision-making in complex ML models
Courts rejecting algorithms
Lack of legal precedents
Lawyers' poor coding skills
Explainable AI (XAI) exists to:
Make predictions faster
Improve transparency and interpretability of ML models
Remove need for lawyers in litigation
Ensure bias in predictions
Quantitative Legal Prediction (QLP) refers to:
Automating legislation
Empirical forecasting of legal outcomes
Eliminating precedent reliance
Teaching lawyers coding
A known example of QLP is:
Lex Machina's IP analytics platform
Law school student ranking
International law drafting
Supreme Court appointments
Research shows predictive algorithms:
Often outperform experts in certain legal contexts
Are less accurate than lawyers
Match lawyers equally
Cannot process case law
One advantage of predictive analytics in law is:
Removing client-lawyer interaction
Faster and more cost-effective analysis
Eliminating human strategy
Creating global uniform law
Adoption of QLP globally is slowed by:
Lack of tech infrastructure and cultural acceptance
Too many predictive tools
Client disinterest in data
Lack of case law
High bias typically results in:
Overfitting
Underfitting
Perfect predictions
Balanced outcomes
Precision measures:
True positives over predicted positives
True positives over actual positives
False negatives over true negatives
Cases filed over cases decided
Recall measures:
True positives over actual positives
True negatives over false negatives
Predicted positives over actual positives
Accuracy of clustering
Curse of dimensionality means:
Too many features reduce model effectiveness
Too few features
Lack of training data
Judicial overload
Balancing bias and variance aims at:
Optimal model performance
Removing dimensionality
Full accuracy
Predicting without data
Overfitting occurs when a model:
Learns noise along with patterns
Generalizes well to new data
Ignores training data
Is too simple
Underfitting occurs when a model:
Captures all noise
Is too simple to capture data complexity
Performs perfectly
Has zero error
Cross-validation is mainly used to:
Prevent bias
Assess generalizability of a model
Automate predictions
Improve case law analysis
A model that performs well on training but poorly on test data is:
Overfit
Underfit
Balanced
Unbiased
Cross-validation improves:
Data bias
Model stability and reliability
Human interpretation
Ethical review
Logistic regression is best for:
Predicting binary outcomes
Predicting continuous values
Grouping clusters
Dimensionality reduction
Maximum likelihood estimation finds parameters that:
Minimize error
Maximize probability of observing data
Eliminate variance
Automate law drafting
In legal prediction, logistic regression could predict:
Whether a judge rules in favor of plaintiff or defendant
Trial duration
Case filing dates
Type of precedent applied
A logistic regression output is:
Probability between 0 and 1
Continuous value
Category clustering
Raw likelihood
Maximum likelihood is important because:
It provides most probable parameter estimates
It avoids model bias
It eliminates overfitting
It standardizes global laws
KNN classifies data by:
Decision trees
Similarity to nearest neighbors
Regression coefficients
Random clustering
Naive Bayes assumes:
Features are dependent
Features are independent
Features are irrelevant
Labels determine features
A strength of KNN is:
Simplicity and non-parametric nature
Perfect interpretability
Infinite scalability
Zero data requirement
A weakness of Naive Bayes is:
Poor with correlated features
High variance
Overfitting
High dimensionality always
In contract classification, Naive Bayes would:
Predict contract type using clause frequencies
Predict damages
Forecast trial length
Clean data
Decision trees work by:
Splitting data into branches using features
Using regression coefficients
Random sampling
Black-box deep models
A key advantage of decision trees is:
Interpretability
Low variance
Elimination of data preprocessing
Perfect predictions
Overfitting in decision trees is reduced by:
Pruning
Expanding
Increasing depth
Ignoring features
Binary classification in trees involves:
Two possible outcomes per case
Unlimited outcomes
Regression models
Probabilistic clustering
A weakness of decision trees is:
Instability to small data changes
Perfect stability
Lack of visual clarity
Complexity
Ensemble methods combine:
Multiple models for better performance
Raw data directly
Features without preprocessing
Laws and statutes
Random Forest reduces variance by:
Averaging predictions from multiple trees
Increasing depth of one tree
Using regression instead
Ignoring features
Bagging in ML is:
Bootstrap aggregation
Bias elimination
Regression method
Dimensionality reduction
Boosting differs by:
Combining weak learners sequentially
Randomizing features
Only using one strong learner
Removing variance
Random Forests are widely used because:
They balance accuracy and interpretability
They eliminate data bias
They require no computation
They always outperform neural networks
K-Means clustering works by:
Assigning points to the nearest centroid
Building decision trees
Predicting binary outcomes
Creating regression lines
The main input needed for K-Means is:
Number of clusters (k)
Regression coefficient
Decision boundaries
Class labels
Hierarchical clustering differs because:
It creates a tree-like structure (dendrogram)
It requires predefined k
It cannot group data
It is always supervised
In legal analytics, clustering can help:
Group similar cases or contracts
Decide trial outcomes
Replace lawyers
Draft laws automatically
A limitation of K-Means is:
Sensitivity to initial centroid placement
It cannot run on computers
It only works on text
It ignores numeric data
Data visualization helps legal professionals:
Identify patterns and communicate insights clearly
Draft contracts
File lawsuits faster
Reduce statutes
A popular R package for visualization is:
ggplot2
numpy
pandas
keras
Good legal data visualization should be:
Clear, accurate, and tailored to audience
Decorative only
Data-heavy with no interpretation
Always in 3D
Visualization in legal analytics may include:
Case trends by year
Lawyer's handwriting
Number of judges per district
Courtroom seating charts
One advantage of visualization is:
Makes complex data accessible and interpretable
Increases data storage
Guarantees perfect predictions
Reduces need for data cleaning
Preprocessing in R often uses:
dplyr for filtering, selecting, and mutating data
keras for clustering
Excel only
SQL exclusively
Removing duplicates from legal data is:
Preprocessing step
Visualization step
Regression step
Prediction step
Normalizing contract dates ensures:
Standard format across datasets
Better legal argument
Judge's fairness
Faster trials
Tokenizing legal text is part of:
Data cleaning
Visualization
Regression
Overfitting
A core benefit of preprocessing is:
Increases accuracy of ML models
Removes the need for models
Eliminates case law
Bypasses ethics concerns
Network analysis studies:
Relationships and connections between entities
Regression coefficients
Decision tree splits
Random forests
In law, network analysis can map:
Judicial citation networks
Criminal handwriting styles
Lawyer salaries
Number of trial rooms
A node in network analysis could represent:
A judge, case, or statute
A regression line
An error term
A cluster centroid only
An edge in legal networks represents:
Relationship such as citations or references
Outcome of case
Model bias
Lawyer's income
Visualizing legal networks helps:
Detect influential cases/judges
Automate contract drafting
Eliminate all precedents
Remove need for lawyers
NLP in law is mainly used for:
Analyzing legal text at scale
Predicting judge salaries
Drafting new statutes
Filing cases automatically
Tokenization in NLP is:
Splitting text into words or phrases
Building regression models
Clustering outcomes
Cleaning data duplicates
Sentiment analysis of court opinions can:
Detect tone or bias in judgments
Replace judicial reasoning
Predict damages directly
Remove case citations
Named Entity Recognition (NER) in law identifies:
Parties, dates, clauses in documents
Regression outputs
Judge's accuracy
Random clusters
A limitation of NLP in law is:
Ambiguity and complexity of legal language
Data always structured
No text data in law
Judges rejecting text
Contract analytics tools can:
Flag non-standard clauses automatically
Write laws
Predict trial duration
Eliminate law firms
SEC data analysis in law often focuses on:
Corporate filings & compliance monitoring
Civil case outcomes
Courtroom seating
Drafting statutes
Judicial prediction tools analyze:
Judges' past rulings and behavior
Lawyers' handwriting
Court architecture
Jury size
Regulatory analytics helps:
Anticipate compliance risks
Draft contracts manually
Predict personal income
Avoid visualizations
Sentiment analysis in legal analytics can:
Gauge public or judicial attitude from texts
Predict numeric damages
Remove precedents
Automate law drafting
Contract review platforms like Luminance use:
Machine Learning + NLP
Only human reading
Only Excel
Only clustering
SEC filing analysis supports:
Investor protection and fraud detection
Criminal handwriting analysis
Drafting constitutions
Jury training
Predictive justice platforms can:
Forecast outcomes with probability scores
Guarantee wins in all cases
Replace courts entirely
Eliminate appeals
A company can use litigation trend analysis to:
Anticipate rising claims and adjust policies
Increase staff salaries
Build new courthouses
Replace HR
Applied analytics in law firms helps:
Resource allocation and client advisory
Erase statutes
Remove precedent law
Automate judgments
SVM works by:
Finding the optimal hyperplane separating classes
Clustering similar items
Regression only
Decision trees
A strength of SVM is:
Effective in high-dimensional spaces
Perfect interpretability
No preprocessing required
Always unsupervised
Kernel trick in SVM allows:
Non-linear separation by mapping data to higher dimensions
Eliminating features
Contract drafting
Removing cases
A limitation of SVM is:
Computationally intensive on large datasets
Always inaccurate
Cannot classify
Only works on text
In legal analytics, SVM could classify:
Case outcomes based on multiple features
Judge's handwriting
Trial dates
Courtroom locations
EM algorithm is used for:
Finding parameters in models with hidden variables
Clustering judges manually
Visualizing contracts
Cleaning text
EM is often applied in:
Mixture models like Gaussian Mixtures
Linear regression
Decision trees
Random forests
EM works through:
Alternating expectation and maximization steps
Only regression steps
One-time parameter estimation
Tokenization
In law, EM can help with:
Modeling hidden patterns in contract data
Predicting trial locations
Writing statutes
Clustering lawyers manually
A drawback of EM is:
Can converge to local optima
Always finds global optimum
No iterative process
Eliminates datasets
Neural networks are inspired by:
Biological neurons in the brain
Court hierarchies
Legal statutes
Random forests
A perceptron is:
Basic unit of a neural network
A clustering centroid
A regression line
A law article
Deep learning refers to:
Neural networks with many hidden layers
Only decision trees
Regression only
Random clustering
In law, neural networks can be applied to:
Analyzing large volumes of case text
Writing laws
Drafting judgments directly
Filing lawsuits
A weakness of neural networks is:
Interpretability ('black box' problem)
Perfect clarity
No training needed
Always linear
MLaaS stands for:
Machine Learning as a Service
Machine Learning and Algorithmic Statistics
Machine Learning Against Systems
Machine Law at Scale
A benefit of MLaaS is:
Scalability and reduced infrastructure costs
No need for lawyers
No ethical concerns
Always interpretable
Cloud-based ML platforms support:
On-demand training and deployment of models
Only Excel-based models
Manual research
Handwritten judgments
Future trends in legal tech include:
Integration of AI with legal workflows
Elimination of law schools
Manual-only research
Removal of precedents
The shifting economics of legal analytics implies:
Data-driven firms gain competitive advantage
All firms must charge higher fees
Clients ignore predictions
Technology has no role
