Financial & AI Glossary

10-K Filing

The comprehensive annual report that public companies file with the SEC, containing audited financials, management discussion, risk factors, and business overview.

10-Q Filing

The quarterly SEC filing containing unaudited financial statements and management discussion for each of the first three quarters of a company's fiscal year.

8-K Filing

The SEC's current report form for material events, which companies must file within four business days of any significant corporate development.

Alpha

The excess return of an investment above what would be predicted by its market exposure, representing genuine skill or informational edge.

Apache Iceberg

An open table format that brings ACID transactions, schema evolution, and time travel to large-scale analytical datasets stored in cloud object storage.

Apache Polaris (Iceberg Catalog)

An open-source implementation of the Iceberg REST Catalog specification that enables multiple compute engines to read and write the same Iceberg tables through a standard API.

Attention Mechanism

The core operation in transformer models that lets each token weight how much it should focus on every other token in the sequence.

Beta

A measure of how sensitively an asset's price moves in relation to a benchmark, typically the overall market.

BigQuery Analytics Hub

Google Cloud's data sharing marketplace that lets publishers list datasets and subscribers access them as native BigQuery tables without data movement.

Bitemporal Data

A data modeling pattern that tracks two independent time axes — when something happened in the real world and when your system first learned about it — enabling rigorous point-in-time queries.

Book Value

The net asset value of a company as recorded in the balance sheet — total assets minus total liabilities — representing the accounting value attributable to shareholders.

Cosine Similarity

A measure of angular similarity between two vectors, ranging from -1 to 1, used to compare embeddings regardless of their magnitude.

Data Lakehouse

An architecture combining the cheap, scalable storage of a data lake with the ACID transactions, governance, and SQL capabilities of a data warehouse.

Debt-to-Equity Ratio

A leverage metric measuring total debt relative to shareholders' equity, indicating how much of the company's financing comes from creditors versus shareholders.

Dense Retrieval

A retrieval approach that encodes both queries and documents as dense vectors, finding relevant matches via nearest-neighbor search rather than keyword overlap.

Drawdown

The peak-to-trough decline in portfolio value over a specific period, measuring how much you lose from the highest point before recovering.

DuckDB

An in-process analytical SQL engine that runs directly in your Python or R session, capable of querying Parquet and Iceberg files on GCS at high speed without a server.

Earnings Call

The quarterly conference call where public company management presents financial results and takes analyst questions, one of the richest sources of unstructured financial information.

EBITDA

Earnings Before Interest, Taxes, Depreciation, and Amortization — a widely used proxy for operating cash flow that strips out financing and accounting decisions.

Effective Timestamp

The business event time — when a fact became true in the real world, independent of when your data system recorded it.

Embedding

A dense numerical vector that encodes the semantic meaning of text, enabling machines to compare and retrieve content by meaning rather than keywords.

EPS (Earnings Per Share)

The portion of a company's net income allocated to each outstanding share of common stock, the most widely cited measure of corporate profitability.

Factor Model

A framework that decomposes asset returns into contributions from systematic risk factors (market, size, value, momentum, quality) plus an idiosyncratic residual.

Fine-Tuning

Adapting a pretrained model to a specific domain or task by continuing training on a targeted dataset, improving accuracy without training from scratch.

Fiscal Period

A company's defined accounting period (quarter or full year) that may differ from the calendar year, used as the primary temporal key in VectorFin's embedding and signal tables.

Forward-Looking Statement

Management's predictions about future performance — guidance, expectations, and outlooks — protected from securities liability by safe harbor rules if accompanied by risk disclosures.

Free Cash Flow

The cash a company generates from operations after paying for the capital expenditures needed to maintain or grow the business — the truest measure of shareholder value creation.

GARCH

A time-series model that captures volatility clustering — the tendency for large price moves to follow other large moves — by modeling variance as a function of past shocks and past variance.

Gross Margin

The percentage of revenue retained after subtracting the direct costs of producing goods or services, a key measure of pricing power and product economics.

Information Ratio

A measure of a portfolio manager's ability to generate active returns per unit of active risk taken, relative to a benchmark.

Knowledge Timestamp

The system ingestion time — when your data pipeline first recorded a fact, used to prevent look-ahead bias in backtesting by filtering to only data available at a given historical date.

Management Discussion & Analysis (MD&A)

The narrative section of 10-K and 10-Q filings where management explains financial results, discusses material trends, and outlines forward-looking expectations.

Momentum

The empirical tendency for assets that have recently outperformed to continue outperforming, and for underperformers to continue underperforming, over a 3-12 month horizon.

Named Entity Recognition (NER)

An NLP technique that automatically identifies and classifies structured entities — companies, people, dates, dollar amounts, locations — within unstructured text.

Operating Leverage

The degree to which a company's operating income amplifies changes in revenue, determined by the ratio of fixed to variable costs in its cost structure.

Parquet

A columnar binary file format optimized for analytical workloads, offering efficient compression and predicate pushdown for large datasets.

Piotroski F-Score

A 9-point accounting-based score that separates financially strong companies from weak ones using signals of profitability, leverage, and operating efficiency.

Point-in-Time Query

A database query that retrieves the data exactly as it existed at a specific historical moment, used to reconstruct past states without contamination from later information.

Price-to-Earnings Ratio (P/E)

The ratio of a stock's price to its earnings per share, measuring how much investors pay for each dollar of current profits.

Regime Detection

Identifying which of several distinct market states (bull, bear, volatile, sideways) the market or a stock currently occupies, to adapt strategy parameters accordingly.

Retrieval-Augmented Generation (RAG)

A technique that grounds LLM outputs by fetching relevant documents from a knowledge base before generating an answer.

Return on Equity (ROE)

A measure of how efficiently a company uses shareholders' equity to generate profit, calculated as net income divided by shareholders' equity.

Safe Harbor Statement

The legal disclaimer preceding forward-looking statements in earnings calls and SEC filings that limits company liability if predictions about future performance prove inaccurate.

Semantic Search

Search that retrieves results based on conceptual meaning rather than exact keyword matching, powered by embedding similarity.

Sentiment Analysis

The automated classification of text as positive, negative, or neutral (and sometimes more granular) to extract opinion and tone signals at scale.

Sharpe Ratio

A measure of risk-adjusted return that divides excess return above the risk-free rate by the volatility of that return.

Tokenization

The process of splitting text into discrete units (tokens) that a language model can process, typically using subword algorithms like BPE or WordPiece.

Transformer

A neural network architecture built on self-attention that has become the foundation for all major language models, including those used to embed financial text.

Vector Database

A specialized database that stores and queries high-dimensional embedding vectors using approximate nearest-neighbor search.

Working Capital

The difference between current assets and current liabilities, measuring a company's short-term liquidity and operational efficiency.