NLP, Computer Vision & Neural Networks – Key Vocabulary

A comprehensive set of vocabulary flashcards covering machine learning fundamentals, NLP processes, neural-network architecture, and CNN applications presented in the lecture notes.

Machine Learning (ML)

A branch of AI that enables systems to learn from data, make predictions, and improve over time without explicit programming.

Data Ingestion

The first ML step involving gathering and preparing a dataset (text, images, numbers, etc.).

Model Training

Using algorithms to let a machine learn patterns in data by iteratively making and correcting predictions.

Testing and Validation

Evaluating an ML model on separate data to ensure it generalizes to unseen examples.

Deployment (ML)

Placing a validated model into a real-world environment where it continues to make decisions and learn.

Supervised Learning

ML approach that learns from labeled input–output pairs to make predictions.

Unsupervised Learning

ML approach that discovers patterns and relationships in unlabeled data.

Reinforcement Learning

ML paradigm where an agent learns by performing actions and receiving feedback (rewards or penalties).

Personalized Recommendations

ML application that suggests products or content (e.g., Netflix, Amazon) based on user data.

Fraud Detection

Use of ML to identify unusual patterns indicative of fraudulent activity in finance.

Natural Language Processing (NLP)

AI field focused on enabling computers to understand, interpret, and generate human language.

Language Understanding

NLP tasks that comprehend spoken or written language, such as speech recognition and sentiment analysis.

Natural Language Generation (NLG)

Producing coherent human-like text from data, e.g., document summarization or chatbot responses.

Text Preprocessing

Initial NLP stage that cleans and structures raw text for further analysis.

Tokenization

Splitting text into smaller units such as words or sub-words.

Normalization

Standardizing text (e.g., lowercasing, removing punctuation) to reduce variation.

Lemmatization

Reducing words to their dictionary base form (lemma).

Stemming

Truncating words to their root form, often by stripping suffixes.

Stop Word Removal

Eliminating common words (e.g., "the", "is") that add little semantic value.

Part-of-Speech (POS) Tagging

Assigning grammatical categories (noun, verb, adjective, etc.) to each word in a text.

Feature Extraction

Transforming text into numerical representations that ML algorithms can process.

Vectorization

Converting text to numeric vectors, often using Bag-of-Words or TF-IDF.

Bag-of-Words

Vectorization method that counts word occurrences without considering order.

TF-IDF

Term Frequency–Inverse Document Frequency; weights words based on their importance across documents.

Word Embeddings

Dense vector representations capturing semantic relationships between words.

Word2Vec

Neural network–based technique that learns word embeddings from large corpora.

GloVe

Global Vectors for Word Representation; embedding method using word-co-occurrence statistics.

BERT Embeddings

Contextual word representations generated by the BERT transformer model.

Model Building (NLP)

Selecting and training algorithms (Naive Bayes, SVM, neural networks, etc.) for a language task.

Natural Language Understanding (NLU)

NLP component that derives meaning, including syntax and semantic analysis.

Syntax Analysis

Examining grammatical structure of sentences (parsing).

Semantic Analysis

Interpreting context and meaning of sentences.

Text Generation

Creating new coherent text based on learned patterns or prompts.

Summarization

Automatically producing concise versions of longer documents.

Evaluation and Refinement

Measuring NLP model performance (precision, recall) and tuning it to reduce errors.

Precision

Metric indicating the proportion of correct positive predictions out of all positive predictions.

Recall

Metric indicating the proportion of actual positives correctly identified by the model.

Application Integration

Embedding an NLP model into products like chatbots, voice assistants, or analytics tools.

Continuous Learning

Ongoing improvement of a model through new data and user feedback.

Feedback Loop

Mechanism that uses real-world results to update and enhance model performance.

Neural Network

Computational model of interconnected artificial neurons organized in layers to recognize patterns.

Input Layer

First neural-network layer that receives raw data from the outside world.

Hidden Layer

Intermediate neural-network layer(s) that progressively transform input representations.

Output Layer

Final neural-network layer that produces predictions or classifications.

Convolutional Neural Network (CNN)

Deep-learning architecture specialized for grid-like data (images) that learns spatial hierarchies of features.

Image Classification

CNN application that assigns an image to one of several categories.

Object Detection and Localization

CNN task of identifying and drawing bounding boxes around multiple objects within an image.

Facial Recognition

CNN-powered identification or verification of individuals based on facial features.

Medical Image Analysis

Using CNNs to detect anomalies (tumors, fractures) in X-rays, MRIs, etc.

Autonomous Vehicles

Self-driving cars employ CNNs for visual perception of pedestrians, road signs, and other vehicles.

Image Enhancement and Restoration

CNN use in improving image quality, colorizing, or repairing damaged photos.

Augmented and Virtual Reality (AR/VR)

CNNs enable real-time image processing to enrich AR/VR experiences.

Robotics (Vision)

Robots employ CNNs for object recognition and navigation in their environments.

Computer Vision

AI field that trains computers to interpret and understand the visual world, often leveraging CNNs.