“Cloze Procedure”: A New Tool for Measuring Readability

By W. L. Taylor. 1953

Subfields: Natural Language Processing, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

In Evaluation & Benchmarks

Citations in the corpus

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding (2019, methodology): “model” (MLM), inspired by the Cloze task (Taylor, 1953).”
Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer (2019, background): “However, it has recently been shown that “denoising” objectives [Devlin et al., 2018; Yang et al., 2019; Taylor, 1953] produce better performance, and as a result they have quickly become standard.”
Adversarial Training for Large Neural Language Models (2020, background): “Essentially, it is a cloze task (Taylor, 1953), where the training objective is the crossentropy loss between the original tokens and the predicted ones.”
Cross-lingual Language Model Pretraining (2019, methodology): “We also consider the masked language modeling (MLM) objective of Devlin et al. (2018), also known as the Cloze task (Taylor, 1953).”
Unified Language Model Pre-training for Natural Language Understanding and Generation (2019, methodology): “In particular, we design a set of cloze tasks (Taylor, 1953) for language models in Table 2, where a masked word is predicted based on its context.”
Teaching Machines to Read and Comprehend (2015, neutral): “Our results indicate the that the neural models achieve a higher accuracy, and do so without any specific encoding of the document or query structure.”
ERNIE: Enhanced Representation through Knowledge Integration (2019, methodology): “To verify ERNIE’s knowledge learning ability, We use several Cloze test samples (Taylor, 1953) to examine the model.”
Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity (2021, methodology): “we use a masked language modeling task (Taylor, 1953; Fedus et al., 2018; Devlin et al., 2018) where the model is trained to predict missing tokens.”
Evaluating the Cross-Lingual Effectiveness of Massively Multilingual Neural Machine Translation (2019, background): “…CWR More recent work has shown that CWRs obtained using unsupervised generative pre-training techniques such as language modeling or cloze task (Taylor, 1953) have led to state-of-the-art results beyond what was achieved with traditional word type representations on many monolingual NLP tasks…”
A Corpus and Cloze Evaluation for Deeper Understanding of Commonsense Stories (2016, methodology): “The cloze task (Taylor, 1953) is used to evaluate a human (or a system) for language understanding by deleting a random word from a sentence and having a human ﬁll in the blank.”
Latent Retrieval for Weakly Supervised Open Domain Question Answering (2019, background): “In the standard Cloze task (Taylor, 1953), the goal is to predict masked-out text based on its context.”
Learning Recurrent Span Representations for Extractive Question Answering (2016, background): “Machine learned extractive question answering systems, such as the one presented here, learn a predictor function f ( q , p ) → a from a training dataset of (cid:104) q , p , a (cid:105) triples.”
Improving and Simplifying Pattern Exploiting Training (2021, background): “A cloze task is a problem where certain parts of a text are removed, and the goal is to replace the missing portion based on the context (Taylor, 1953).”
Foundation Posteriors for Approximate Probabilistic Inference (2022, neutral): “The MLM objective is a word prediction task [56].”
What Language Model Architecture and Pretraining Objective Work Best for Zero-Shot Generalization? (2022, methodology): “language modeling (MLM) objective (Taylor, 1953; Devlin et al., 2018), such as span corruption (Raffel et al.”
Iterative Alternating Neural Attention for Machine Reading (2016, background): “Cloze-style queries (Taylor, 1953) are created by deleting a particular word in a natural-language statement.”