Precision and Recall
Two complementary metrics for evaluating information retrieval systems: precision measures accuracy of returned results, recall measures completeness.
Precision = True Positives / (True Positives + False Positives) — of the items returned, what fraction were actually relevant? Recall = True Positives / (True Positives + False Negatives) — of all relevant items that exist, what fraction were returned? There is typically a trade-off: increasing recall (returning more results) reduces precision; increasing precision (returning only confident matches) reduces recall.
For document intelligence in legal and compliance contexts, the relative importance of precision and recall depends on the use case. In M&A due diligence, missing a change-of-control clause (false negative) may be far more costly than flagging an irrelevant clause for human review (false positive) — suggesting high-recall retrieval is appropriate. In executive summaries, precision matters more — including irrelevant information reduces trust and increases review burden. The F1 score (harmonic mean of precision and recall) provides a single balanced metric when both matter equally.
More ai/ml Terms
Retrieval-Augmented Generation (RAG)
An AI architecture that combines information retrieval with text generation to produce answers grounded in source documents.
Vector Embedding
A numerical representation of text as a high-dimensional vector, enabling semantic similarity comparisons between passages.
BM25
A probabilistic keyword-ranking algorithm that scores documents by term frequency and inverse document frequency.
Chunking
The process of splitting large documents into smaller, overlapping segments optimized for retrieval and embedding.
Hallucination
When an AI model generates plausible-sounding but factually incorrect or fabricated information.
Large Language Model (LLM)
A neural network trained on massive text corpora that can understand and generate human language.
Analyze Documents Related to Precision and Recall
Upload any document and get AI-powered analysis with verifiable citations.
Start Free