ChatGPT is currently performing reasonably well in the design of entries for a specific type of dictionary, for the moment of the English language, shows Prof Robert Lew of the Faculty of English. His study on this subject was recently published in Humanities and Social Sciences Communications!
The first dictionary, based on Artificial Intelligence, was the COBUILD dictionary, using, among other things, an innovative definition format to illustrate the use of a word in context.
The Polish equivalent of the COBUILD dictionary is Mirosław Bańko's 'Inny słownik języka polskiego'. Definitions contained there are simple to understand and conversational.
Historically, lexicography was an art of intuition or relied on copying ideas from older dictionaries. Yet, at the end of the 20th century, under the influence of language corpora, leading lexicography began to be an empirical procedure strictly grounded in data. A language corpus thus examined the contexts in which words were used most frequently. On this basis, a lexicographic description of the language, including definitions and usage illustrations was created.
It is already apparent that we can partly automate the process of dictionary design. However, it is necessary to have well-developed prompts (instructions for the bot). However, it is at the stage of early experiments for now. And the results are promising for English but not yet for Polish.
So far, we have seen language as something very human, uniquely human, something that nobody will take away from us. And here ChatGPT has been made available, and we have 'proof of the pudding' - the pudding is out there, and everyone can try it. Text generators already exist. They know how to use the language, and nobody can check it out. For English, it already works, but for many languages - so not quite," - explains Prof Robert Lew.
The entire article is available at: https://www.nature.com/articles/s41599-023-02119-6
photo credit: A. Wykrota
