Bag of Words is a lexical representation model where a document is expressed as a collection of its words, disregarding grammar and order. Each word in the vocabulary becomes a feature dimension, and documents are represented by vectors of word ...

Latent Semantic Analysis is a mathematical technique that uses Singular Value Decomposition (SVD) to reveal hidden relationships in large text corpora. Surface Level (BoW/TF-IDF): Words are treated as independent, literal tokens. Latent Level (LSA): Words and documents are mapped into ...

LDA is a Bayesian topic model that uncovers the latent structure of text. Instead of classifying a document into a single category, it treats every document as a mixture of multiple topics. A document might be 60% “machine learning” and ...

A document embedding is a fixed-length vector representation of an entire text — whether a sentence, paragraph, or full page. Lexical models (BoW, TF-IDF) only capture word presence or frequency. Document embeddings encode semantic similarity between texts, allowing machines to ...

A Sequence-to-Sequence (Seq2Seq) model is a neural network architecture designed to transform one sequence into another, such as translating a sentence, summarizing a document, or converting speech into text. Key components: Encoder → Reads the input sequence and compresses it ...

Text classification is built on a pipeline of preprocessing, feature extraction, modeling, and evaluation. The most common features include bag-of-words and TF-IDF, which represent documents as weighted vectors of terms. This process is similar to how information retrieval systems operate: ...

Information Extraction transforms unstructured text into structured forms, enabling downstream reasoning. It includes: Named Entity Recognition (NER): spotting entity mentions. Relationship Extraction (RE): mapping links between entities. Event Extraction: capturing actions and their participants. NER provides the nodes, while RE ...

Machine Translation is the process of converting text in one language into another while preserving meaning, style, and fluency. Unlike a dictionary lookup, MT must navigate: Ambiguity (words with multiple meanings). Grammar and word order differences. Morphological complexity across languages. ...

Text summarization aims to condense content while preserving meaning. Two broad categories exist: Extractive Summarization: Selects important sentences directly from the source text. Abstractive Summarization: Generates new sentences to convey the same meaning in a more concise form. Extractive methods ...

Semantics focuses on how words and sentences convey meaning. But treating queries as static strings often fails in practice. Pragmatics introduces an additional dimension: it asks why a query was made, what assumptions the user and system share, and whether ...