Lexicography is the craft and discipline of writing, editing, and compiling dictionaries. It sounds straightforward, but the work involves far more than listing words and their meanings. Lexicographers make thousands of judgment calls about which words to include, how to define them, what usage examples to provide, and which senses of a word deserve separate entries. The field has shifted dramatically over its history, moving from intuition-driven prescriptivism to data-rich descriptivism, and is now being reshaped again by artificial intelligence.
From Word Lists to Cognitive Dictionaries
The history of English lexicography stretches back to simple glossaries that translated difficult Latin or French words into vernacular English. Over centuries, those word lists grew into full dictionaries, and the guiding philosophy behind them changed several times. Scholars have traced a coherent arc from the earliest “archetype” dictionaries, through prescriptive works that tried to fix “correct” usage, to the great historical dictionaries like the Oxford English Dictionary, and then onward to descriptive dictionaries that aimed to record how people actually use language rather than how they ought to use it.1Lexikos. Towards an Evolutional Chain of English Dictionary Paradigms from the Linguistic Perspective The most recent stage in this chain is the cognitive dictionary, which tries to reflect how the human mind organizes and retrieves word knowledge.
The prescriptive era left a lasting mark on how most people think about dictionaries. Samuel Johnson’s 1755 dictionary and Noah Webster’s American dictionaries were openly judgmental about usage, labeling some forms as vulgar or incorrect. That sensibility still shows up when someone says “that’s not a real word” or insists that a dictionary definition settles an argument about meaning. Modern professional lexicography, though, has moved well past that stance. The dominant approach today is descriptive: a dictionary should document how people use language, including informal speech, slang, and contested usages, rather than legislate what counts as proper.
How Modern Dictionaries Get Made
The single biggest change in practical lexicography over the past few decades has been the rise of corpus linguistics. A corpus is a large, searchable collection of real-world text, sometimes billions of words drawn from books, newspapers, transcripts, web pages, and other sources. Before corpora became standard tools, lexicographers relied heavily on their own reading and on citation slips, handwritten notes recording interesting or illustrative uses of a word. The process was slow, idiosyncratic, and inevitably skewed toward whatever the lexicographer happened to have read.
Corpus tools changed this fundamentally. Software like AntConc and LancsBox lets lexicographers extract statistically significant collocations (words that frequently appear together), concordance lines (a word shown in the context of the sentences surrounding it), and frequency data. These tools influence nearly every stage of dictionary production: selecting which headwords to include, writing definitions that reflect actual usage patterns, choosing example sentences drawn from authentic text rather than invented illustrations, and documenting phrasal patterns.2Advance Social Science Archive Journal. The value of Corpus Linguistics for contemporary dictionary making: From Citations to Collocations The shift from intuition to data means that a modern learner’s dictionary definition of “run,” for instance, reflects not what a single editor thought the word meant but what millions of uses of “run” in context actually show.
This does not make the lexicographer’s job mechanical. Corpus data reveals patterns, but a human still has to interpret ambiguous cases, decide how to split senses, and write prose that a reader can understand. The data keeps the lexicographer honest; the lexicographer keeps the dictionary readable.
Writing Definitions People Can Understand
One of the less visible but most consequential innovations in dictionary-making is the controlled defining vocabulary. The idea, pioneered by the Longman Dictionary of Contemporary English, is to write every definition using only a restricted set of roughly 2,000 common words. If a definition needs a word outside that set, it must be defined elsewhere in the dictionary using the same limited vocabulary.
The payoff is real. Studies comparing learner dictionaries found that users judged definitions written with a controlled vocabulary to be more helpful and easier to understand than definitions in competing dictionaries that did not impose the same constraint. The controlled vocabulary was also shown to produce definitions that were both more comprehensible and sufficiently accurate, lightening the learning burden for foreign-language students.3ResearchGate. A Critique of the Controlled Defining Vocabulary in Longman Dictionary of Contemporary English This is not just a convenience feature. For the hundreds of millions of people learning English as a second language, the difference between a definition that uses accessible words and one that sends you into a chain of further lookups can determine whether the dictionary is a useful tool or a frustrating loop.
The trade-off is that restricting your vocabulary can sometimes make definitions vaguer or more roundabout. Defining a technical concept without using any technical terms forces a kind of paraphrasing that may lose precision. Lexicographers working with controlled vocabularies constantly negotiate between clarity for the learner and accuracy for the subject. Some critics argue the restriction goes too far, others that it does not go far enough. But the evidence that it helps comprehension has been consistent enough that the approach is now standard in major learner dictionaries.
Dictionaries as Cultural Artifacts
Dictionaries feel authoritative and neutral. They sit on reference shelves, get cited in court cases, and are treated as arbiters of meaning. But this apparent neutrality masks the fact that every dictionary is produced by people who live in a particular time, place, and culture, and whose perspectives shape what gets recorded and how.
Dictionaries are, in an important sense, partisan systems of ordering words and meanings. They may aim to be universal, but they inevitably emerge from and respond to social moments from particular perspectives. Those perspectives can celebrate or marginalize cultural groups, legitimate or suppress certain languages, and facilitate either social mobility or discrimination.4Cambridge University Press. Dictionaries and cultural politics A dictionary might present its subjective editorial style as objective and universal, or it might highlight its cultural positionality openly, and either stance can serve political or commercial purposes.
You can see this playing out in ongoing public debates about dictionary entries for words related to gender, race, and identity. When a dictionary adds a new sense of a word or updates a definition, it is not merely recording a linguistic change in a vacuum. It is making a decision about what counts as standard, who gets represented, and which usage communities matter enough to warrant documentation. The firestorms that sometimes erupt over dictionary updates are not really about linguistics. They are about who gets to shape the shared reference point that a dictionary represents.
This also applies across languages. Bilingual dictionaries, for example, do not just translate words; they make equivalence judgments between concepts that may not map neatly from one culture to another. A word in one language might carry connotations, social registers, or emotional weight that no single word in the target language captures. The lexicographer’s choice of translation equivalent is an editorial act, not a neutral mirror.
How People Actually Look Up Words Now
The shift from print to digital has changed not just how dictionaries are made but how they are used. Research into dictionary search behavior shows that students overwhelmingly rely on online dictionaries, and their primary motivation is comprehension rather than production. That is, they look words up to understand something they are reading, not to find the right word for something they are writing.5Journal of English Language Teaching. Analyzing Dictionary Search Preferences: A Study of User Behaviour
This matters for dictionary design. If most lookups are driven by reading comprehension, the dictionary needs to prioritize clear, accessible definitions and good example sentences over extensive production-oriented features like usage notes on near-synonyms or detailed grammatical coding. Print dictionaries tried to serve all purposes in a single volume because they had to. Digital dictionaries can, in theory, tailor the information they present based on whether someone is trying to understand a word or use it. Few have fully exploited this possibility yet, but the data about how people actually search points in that direction.
The speed of digital lookup has also changed reading habits. When you can tap a word and see its definition in a second, the threshold for looking something up drops. Words that a print-dictionary user might have guessed from context now get looked up explicitly. Whether this is making people better readers or more dependent on external tools is an open question, but the volume of dictionary consultations has almost certainly increased.
Dictionary Use and Vocabulary Learning
For language learners, the question of when and how to use a dictionary is more nuanced than it might seem. Simply looking up every unknown word does not automatically produce lasting vocabulary knowledge. Research on reading and vocabulary acquisition has found that the most effective approach combines inferencing (guessing a word’s meaning from context) with dictionary consultation. Using both strategies together produces the greatest gains in both reading comprehension and long-term vocabulary retention.6Canadian Journal of Applied Linguistics. The Role of Consulting a Dictionary in Reading and Vocabulary Learning
There is a cost, though. Consulting a dictionary slows down the reading process. Think-aloud studies, where learners narrate their thought process while reading, reveal that strategic dictionary use contributes to several cognitive processes that support vocabulary acquisition: paying attention to the connection between a word’s form and its meaning, rehearsing the word in a way that supports long-term memory storage, and building associations with other knowledge the learner already has.7Canadian Journal of Applied Linguistics. The Role of Consulting a Dictionary in Reading and Vocabulary Learning The slowdown, in other words, is partly the point. The mental effort of pausing, looking up a word, and integrating the new meaning into the sentence you are reading is part of what makes the learning stick.
The practical takeaway for learners is counterintuitive: you should try to guess a word’s meaning before you look it up, and you should not feel guilty about the time the lookup takes. The combination of guessing and then confirming or correcting your guess is what produces the deepest learning. Skipping the dictionary entirely is faster but less effective; looking up every word without trying to infer first is also less effective. The sweet spot is in between.
Artificial Intelligence and Dictionary-Making
The latest upheaval in lexicography comes from large language models like ChatGPT and their ability to generate text that resembles dictionary entries. Recent studies have shown that generative AI can produce good-quality dictionary entries, and in some comparisons, a free version of ChatGPT outperformed an established learner dictionary on both comprehension and production tasks with university students.8Humanities and Social Sciences Communications. Dictionaries and lexicography in the AI era This has prompted serious discussion about whether the traditional dictionary format, with its carefully crafted entries and rigid structure, may eventually be supplanted by conversational AI that can answer lexical questions on the fly.
The argument for AI is straightforward: if you can ask a chatbot “what does ‘recalcitrant’ mean?” and get a clear, contextually appropriate answer, why would you navigate a dictionary interface? The argument against is equally clear: AI models can hallucinate (produce confident-sounding but incorrect information), they lack the editorial accountability of a published dictionary, and their answers can vary depending on how you phrase the question. A dictionary entry that was wrong would be corrected in the next edition. An AI-generated definition that is wrong on Tuesday might be right on Wednesday and wrong again on Thursday, with no changelog.
There are also applications where AI assists lexicographers rather than replacing them. One recent project used large language models to automatically compile a lexicon of Pre-Qin Chinese philosophy, handling tasks like identifying terms, classifying them by philosophical school, generating definitions, and translating context passages. The system produced a structured term database and an interactive interface that supports an end-to-end workflow from raw text to dictionary display.9npj heritage science. Automatic compilation of a Pre-Qin philosophy lexicon via large language models For specialized or historical lexicography, where the volume of text to process is enormous and expert human labor is scarce, this kind of AI-assisted pipeline could be transformative.
The field is still working out where the line between “AI-assisted” and “AI-generated” should fall. Most lexicographers see AI as a powerful drafting and analysis tool that still needs human oversight, especially for sensitive definitions, culturally loaded terms, and edge cases where corpus data is thin. But the pace of improvement in language models is fast enough that the boundaries are shifting year by year.
Copyright and the Legal Side of Dictionary Work
One dimension of lexicography that rarely gets public attention is intellectual property. Lexicographers occupy a peculiar dual position: they are both creators of copyrightable products and users of copyrighted material. A dictionary is an original work of authorship (the definitions, the organizational structure, the selection and arrangement of entries), but it is also built from the language of others (quotations, example sentences, source texts consulted during research).10Lexikos. Lexicography, terminography and copyright
This creates some thorny legal questions. Can you copyright a single dictionary definition? Generally, yes, if the definition reflects enough original expression, but a short, utilitarian definition of a common word may not clear the threshold of originality. Can you copy definitions from one dictionary into another? Almost certainly not without permission, though paraphrasing to the point of independent expression is a gray area. What about example sentences drawn from copyrighted novels or articles? Fair use or fair dealing provisions may apply, but the boundaries depend on jurisdiction and on how much of the original text is reproduced.
The digital era has made these questions more urgent. When dictionaries were heavy books that took years to produce, copying was slow and obvious. Now, dictionary content can be scraped, repurposed, and redistributed almost instantly. Online dictionaries have to think about licensing, digital rights management, and the terms under which their content can be accessed by third-party apps or AI training pipelines. The rise of user-generated dictionary content, wiki-style collaborative projects, and AI-generated definitions adds further complexity. If an AI produces a definition that closely resembles an existing copyrighted definition, who is liable? These questions are still being worked out in courts and legislatures around the world.
Specialized and Minority-Language Lexicography
Most of the public conversation about dictionaries centers on major languages with large speaker populations and well-funded publishing ecosystems. But some of the most important lexicographic work happens at the margins: documenting endangered languages, creating first-ever dictionaries for languages that have only been spoken and never written down, and building specialized terminological resources for fields like law, medicine, or engineering.
For endangered languages, a dictionary can be an act of cultural preservation. When a language has only a few hundred elderly speakers left, the lexicographer’s work takes on an urgency that commercial English dictionary-making does not share. These projects often involve deep collaboration with native speaker communities, and they raise their own ethical questions about ownership, access, and who gets to decide how a language is represented in print.
Specialized dictionaries, meanwhile, face a different challenge: their users need extreme precision, not accessibility. A medical dictionary that simplifies a term’s definition for readability may actually become less useful to the clinicians who need it. A legal dictionary that smooths over jurisdictional differences in how a term is used may mislead a practitioner in a specific court system. The lexicographic principles are the same as in general dictionaries, but the priorities shift. Controlled defining vocabularies, for instance, would be counterproductive in a dictionary whose audience consists entirely of domain experts who need unambiguous technical language.
These specialized corners of the field rarely make headlines, but they collectively represent a huge volume of lexicographic labor. Every time a new technical standard is published, every time an international organization needs consistent terminology across six official languages, every time a translator encounters a term with no established equivalent in the target language, lexicographic work is happening, even if nobody calls it that.