In grammar education, we propose a new educational paradigm that expands thinking through questioning, fostering both teacher expertise and student inquiry. We view AI as a learning partner, not a machine that simply provides answers.
Artificial Intelligence Encyclopedia.
You can find the artificial intelligence knowledge you need at aiseries.oopy.io.

So, how is the field of grammar education responding to the emergence of AI amidst this massive wave of change? Unfortunately, compared to other fields, its response has been relatively muted.
Nearly three years have passed since ChatGPIT was introduced, but domestic research on AI-based grammar education for native speakers is still scarce.
This silence is even more striking compared to the active discussion about introducing AI in the fields of writing and literature education, not to mention math and English, even within the Korean language curriculum.

Why is this? There are three main reasons.
First, it may be because there is still no clear picture of what role AI can play in grammar education.
Second, it may be premature to use AI in education due to its imperfect performance in Korean grammar.
Third, the confusion may stem from the fact that AI does not explicitly learn grammar rules but rather uses language patterns from data, which is a very different approach from the way grammar education has been conducted so far.
--- From “01_“AI and Grammar Education Trends””

Piantadosi also suggests that AI may have a sense of grammar.
When he showed GPT-3.5 the famous example Chomsky gave, “Colorless green ideas sleep furiously,” the AI ​​responded that the sentence was grammatically correct but semantically strange.
On the other hand, when “Green sleep furiously colorless ideas” was presented, it was pointed out that it was grammatically awkward.
This suggests that, contrary to Chomsky's claims, AI has internalized the ability to judge the 'grammaticality' of a sentence independently of its meaning.
--- From "03_“AI Grammar and Human Grammar”"

Above all, the involvement of a professional grammar teacher is essential.
Those who best understand the real-world context of grammar education are teachers in the field.
This is because they are the 'field experts' who know best which grammar concepts students frequently misunderstand, which explanation methods are effective, and when and how feedback should be provided.
Teachers are also those who can transform the same grammar knowledge into diverse "pedagogical content knowledge (PCK)" tailored to the learner's situation. For AI to become a competent tutor, it must learn from high-quality data directly created by these teachers.

But teachers alone are not enough.
Collaboration with data experts is essential for teachers' expertise and know-how to be converted into "data" that AI can learn from.
Since the explanations or conversation materials provided by the teacher are in natural language, they must be converted into a structure that AI can understand.
This process requires technical work such as attaching part-of-speech information to sentences (tagging), analyzing syntactic structure, and classifying types of feedback.
Additionally, designing the hierarchy and relationships of grammatical concepts in the form of ontology also requires a technical understanding of knowledge representation languages ​​(e.g., RDF, OWL).
--- From “06_“The Importance of High-Quality Data””

The view on word meaning according to neostructuralism is summarized in the distributional hypothesis.
This can be summarized in the words of linguist John Firth: “You shall know a word by the company it keeps.”
That is, the meaning of a word is determined by the statistical distribution information of which words the word mainly appears with in a large text.
For example, 'warm' tends to appear with 'coffee', 'weather', and 'heart', and 'cold' tends to appear with 'ice', 'wind', and 'expression'.
By statistically analyzing the patterns of words that appear together like this, we can objectively define the semantic difference between 'warm' and 'cold'.
This distribution hypothesis becomes the principle of natural language processing technology in modern artificial intelligence.
Word embedding models such as Word2Vec are representative examples of mathematical implementations of the distributional hypothesis.

--- From “09_“Changes in the Content of Grammar Education””