Steve, It would be great if I could somehow prime the Chat feature with a starting request such as “Por versus Para” or “Indirect pronouns” or “Present subjunctive”, which would prime the Chatbot to ASK leading questions to encourage me to try to respond according to the theme I’ve selected. I guess the same feature could prime the Chatbot to ask me questions eliciting, say, answers concerning dining at a restaurant or visiting Paris, if that was what the user felt they most wanted to stress.
As it is, the Chatbot keeps trying to get the topic OFF whatever I am trying to direct the topic toward, so it is a bit of a “wrestling match”…which I always win, since I just keep ignoring the shallow questions it asks and babbling on (actually talking…because I am using an Apple tablet with dictation turned on, and turning to the translate feature if my Spanish is inadequate to say what I want to say in the way I want to say it).
Anyway, I am doing this as a kludge, pretty effectively. So why not just feed the Chatbot to dwell on a LEADING question or theme. That would encourage me to stress the topics it is primed to stress. (E.g., stress preterite by asking questions and ASKING for responses with the preterite versus imperfect). If that is too hard, at least let me say “Ask about what I did in the past day”, and get the chatbot to stay on theme.
I have no idea how feasible this is with the current AI chatbot, but it actually does seem to good-naturedly bend its responses to my heavy-handed onslaught! So why couldn’t it be primed with the user submitting a TOPIC or GRAMMATICAL BIAS as a lead-in to the session?
A related thing would be for me to supply the Chatbot with a list of phrases I want to stress and having it, then, make those part of the ‘theme’ and maybe do some sort of ranking. In a worst case, it would be cool if the chatbot could ask things as heavy handed as “How would you say, ‘When I was I child I went to the store’?” Again, I have no idea if the existing chatbot can be guided in some way similar to this; but I presume next year, and the year after, are going to see some pretty dramatic upgrades in sophistication.
Steve, I just thought of an alternate way to get me what I am looking for: Can the Chatbot be set up to allow the user to submit an ESSAY – read, say, a max 300 word text input buffer – as the heart of the session? The user would then be able to get the essay corrections, on the topic and using the grammar/tense of their choice, all in one place. This could even be as an alternative to an actual chat, if the cost of correcting so many words sentences is too high.
That way I could submit a soliloquy talking about what I want talk about and get integrated feedback aimed toward the things I want to talk about in the way I want to talk about them. Yes, this is a bit different than just having an improptu “conversation”, but it would arguably be a much better LEARNING opportunity, since the full context of what I said originally and how I could maybe express it more gramatically would be there for me to compare and review, side-by-side.
It seems to me it only requires a larger buffer input, for the user to supply in advance of starting the chat. And, as I said, if it is an issue of limiting use of the AI to keep it below some ceiling per user session, then the amount of follow-on interaction with the chatbot could be capped as necessary.
I am SOMEWHAT doing the same thing, already, it is just that my “essay” tends to be badly organized since it has to be constructed in pieces, on the fly. But most of my responses / inputs are pretty wide ranging anyway. So it SEEMS to me that reading first from a larger input buffer would just be in place of my opening “Hola…” greeting.
A lot of my “responses” back to the chatbot are along the lines of “why did you say it that way versus this way? Is one better than the other?” Or “can you give me some sample sentences using this pattern to practice with?”
Anyway, the opening context for the whole conversation could be set in the input buffer if the user wishes that. If the user doesn’t put anything in the input buffer then it the conversation would default to the same sort of often inane “Hello, how are you today…” opening that otherwise currently happens (as it does in real-world conversations, of course…where only the trivia generally gets covered before the participants run out of time;-)