That Online Corpus of Founding Era American English seems like a pretty cool database. This is five years old (pre ChatGPT) and seems to have relied on manual search (which itself seems like a vast improvement). I wonder whether large language models are being built to assimilate the entire dataset to answer questions about “original meaning” nowadays and how close to useable they are. It would be even more compelling to have longitudinal versions that can identify when changes in meaning occurred. “Based on all existing written words, it didn’t mean X at that time and that meaning first appeard 60 years later.” Newspapers and legal rulings/documents seem like relatively convincing data sources that have been well curated and relevant to the task. Particularly since SCOTUS post-Scalia has become even more insistent about original meaning. I don’t think it works well post-hoc but it will be interesting for these things to be interpreted when presented as arguments in new cases.
That Online Corpus of Founding Era American English seems like a pretty cool database. This is five years old (pre ChatGPT) and seems to have relied on manual search (which itself seems like a vast improvement). I wonder whether large language models are being built to assimilate the entire dataset to answer questions about “original meaning” nowadays and how close to useable they are. It would be even more compelling to have longitudinal versions that can identify when changes in meaning occurred. “Based on all existing written words, it didn’t mean X at that time and that meaning first appeard 60 years later.” Newspapers and legal rulings/documents seem like relatively convincing data sources that have been well curated and relevant to the task. Particularly since SCOTUS post-Scalia has become even more insistent about original meaning. I don’t think it works well post-hoc but it will be interesting for these things to be interpreted when presented as arguments in new cases.