Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Misspellings in Natural Language Processing: A survey

Gianluca Sperduti, Alejandro Moreo

TL;DR

This survey examines misspellings as a pervasive challenge in NLP, tracing historical shifts from curated text to Web-driven, data-rich settings and highlighting the limits of traditional preprocessing. It catalogs a spectrum of mitigation approaches, including data augmentation, adversarial training, and order-agnostic representations, and surveys datasets, tasks, and benchmarks used to study robustness. The discussion extends to large language models, revealing that even state-of-the-art systems suffer performance drops under misspellings and prompting careful evaluation via instance- and prompt-based tests. The work emphasizes safety and ethical considerations, such as adversarial misspellings for malign content, and advocates for integrating resilience into NLP systems to ensure reliable real-world deployment across domains like MT and content moderation.

Abstract

This survey provides an overview of the challenges of misspellings in natural language processing (NLP). While often unintentional, misspellings have become ubiquitous in digital communication, especially with the proliferation of Web 2.0, user-generated content, and informal text mediums such as social media, blogs, and forums. Even if humans can generally interpret misspelled text, NLP models frequently struggle to handle it: this causes a decline in performance in common tasks like text classification and machine translation. In this paper, we reconstruct a history of misspellings as a scientific problem. We then discuss the latest advancements to address the challenge of misspellings in NLP. Main strategies to mitigate the effect of misspellings include data augmentation, double step, character-order agnostic, and tuple-based methods, among others. This survey also examines dedicated data challenges and competitions to spur progress in the field. Critical safety and ethical concerns are also examined, for example, the voluntary use of misspellings to inject malicious messages and hate speech on social networks. Furthermore, the survey explores psycholinguistic perspectives on how humans process misspellings, potentially informing innovative computational techniques for text normalization and representation. Finally, the misspelling-related challenges and opportunities associated with modern large language models are also analyzed, including benchmarks, datasets, and performances of the most prominent language models against misspellings. This survey aims to be an exhaustive resource for researchers seeking to mitigate the impact of misspellings in the rapidly evolving landscape of NLP.

Misspellings in Natural Language Processing: A survey

TL;DR

This survey examines misspellings as a pervasive challenge in NLP, tracing historical shifts from curated text to Web-driven, data-rich settings and highlighting the limits of traditional preprocessing. It catalogs a spectrum of mitigation approaches, including data augmentation, adversarial training, and order-agnostic representations, and surveys datasets, tasks, and benchmarks used to study robustness. The discussion extends to large language models, revealing that even state-of-the-art systems suffer performance drops under misspellings and prompting careful evaluation via instance- and prompt-based tests. The work emphasizes safety and ethical considerations, such as adversarial misspellings for malign content, and advocates for integrating resilience into NLP systems to ensure reliable real-world deployment across domains like MT and content moderation.

Abstract

This survey provides an overview of the challenges of misspellings in natural language processing (NLP). While often unintentional, misspellings have become ubiquitous in digital communication, especially with the proliferation of Web 2.0, user-generated content, and informal text mediums such as social media, blogs, and forums. Even if humans can generally interpret misspelled text, NLP models frequently struggle to handle it: this causes a decline in performance in common tasks like text classification and machine translation. In this paper, we reconstruct a history of misspellings as a scientific problem. We then discuss the latest advancements to address the challenge of misspellings in NLP. Main strategies to mitigate the effect of misspellings include data augmentation, double step, character-order agnostic, and tuple-based methods, among others. This survey also examines dedicated data challenges and competitions to spur progress in the field. Critical safety and ethical concerns are also examined, for example, the voluntary use of misspellings to inject malicious messages and hate speech on social networks. Furthermore, the survey explores psycholinguistic perspectives on how humans process misspellings, potentially informing innovative computational techniques for text normalization and representation. Finally, the misspelling-related challenges and opportunities associated with modern large language models are also analyzed, including benchmarks, datasets, and performances of the most prominent language models against misspellings. This survey aims to be an exhaustive resource for researchers seeking to mitigate the impact of misspellings in the rapidly evolving landscape of NLP.

Paper Structure

This paper contains 32 sections, 4 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. A Brief History of Misspellings
  3. Before 2010: Fewer Data, Less Misspellings
  4. After 2010: The Rise of Social Networks and Deep Learning
  5. What is a Misspelling?
  6. Under the Lens of Linguistics
  7. Under the Lens of NLP
  8. How Serious is the Problem?
  9. The Harm of Synthetic Misspellings
  10. The Harm of Natural Misspellings
  11. Methods
  12. Data Augmentation
  13. Generalized Data Augmentation
  14. Adversarial Training
  15. Character-Order-Agnostic Methods
  16. ...and 17 more sections

Figures (3)

  • Figure 1: Publication trends in NLP papers on misspellings (2004-2024)
  • Figure 2: Conceptualization of an order-agnostic representation for garbled words. Dotted lines denote garbled variants of the original word, on the top. Solid lines denote an order-agnostic representation of a surface form word. If all characters (but the first and last) are represented as a set, then the representation of the original word and the garbled variants coincide.
  • Figure 3: Distribution of methods (left) across tasks (right)