liveBook · Manning

Cloze test

As explained in AI-Powered Search

A Cloze test is a language-comprehension assessment that evaluates whether a model can predict omitted tokens or words from context.

A cloze test evaluates language comprehension by removing a word or token from context and asking the reader or model to predict what belongs in the blank. section 13.3.1

Predicting a missing word

Given “I went to the ________ and bought some vegetables,” people may answer “store,” “market,” or “grocer.” Comparing answers tests how context supports comprehension. section 13.3.1

Longer comprehension questions extend the same idea: a reader must locate and recall information from surrounding text. These assessments provide a foundation for understanding how language models learn from large text collections. section 13.3.1

How AI-Powered Search covers it

How deep it goes
Explains the assessment through classroom examples and connects it to language-model training, without detailing model implementation.
Examples
Uses a missing-word classroom prompt and a passage question asking how far the churchyard stood from the sea.
What it assumes
Assumes reading ability and contextual knowledge, while illustrating the assessment with familiar sentences and passages.
What you can do after
Evaluate language comprehension by examining whether missing words or answers follow from surrounding context.
How it connects
Connects reading comprehension with language models, NLP, contextual knowledge, and Transformer-based language understanding.

This explainer was generated by Manning's knowledge graph from the text of AI-Powered Search. The section links point to the parts of the book it draws on.