2021-03-31
What would you like to see happen in this conversation?
Meaning is not a set., I had a realization, and I would like to discuss if this has any practical application.
What is the "application" of this "application"?
For example, chatbots will be able to do things they couldn't do before.
What kind of "chatbot" is this "chatbot"?
Understand the human story.
What kind of "understanding" is that "understanding"?
We understand the meaning of what people are talking about. We don't yet know what it means to understand the meaning, but there is progress in the processing that deals with meaning.
What is the "story" about?
I wonder, that remains to be seen. Maybe it's the kind of dialogue we're having here, maybe it's not.
Is there anything else about that "story"?
There is no And when a meaning is a point, there is similarity between the meanings, but no inclusion. When a meaning is a set, there is an inclusion. There can be intersection. When a meaning is a Gaussian, the probability of inclusion is required.
What is it that you value?
Hopefully there will be some useful application.
Is there anything else you can tell us about its "application"?
That's what I'm looking for.
I wonder, that remains to be seen. Maybe it's the kind of dialogue we're having here, maybe it's not.
- Where is this "story"?
Between human and computer.
I wonder, that remains to be seen. Maybe it's the kind of dialogue we're having here, maybe it's not.
- Where is this "story"?
It's like using digital communication to flow in text form.
I wonder, that remains to be seen. Maybe it's the kind of dialogue we're having here, maybe it's not.
- How do you know this "story"?
I don't know.
What kind of "understanding" is that "understanding"?
If we find something better, we'll know.
Is there anything else about that "understanding"?
I want to understand.
Hopefully there will be some useful application.
- Where is this "application"?
That's what I'm looking for.
Hopefully there will be some useful application.
- What does this "application" look like?
Hmmm, it's the next step after the stairs, and if you can do that, you can reach even higher.
I wonder, that remains to be seen. Maybe it's the kind of dialogue we're having here, maybe it's not.
- What do you hope will happen regarding this "story"?
For example, with regard to the conversation we are having here, it would be nice if the questions were more appropriate. For example, if we understand that the user is talking about an event, then the question "What happens next?" is appropriate. question is appropriate.
Where does this "story" come from?
If the computer asks the right questions, it will pull from the human.
What kind of "human being" is this "human being"?
Wait a minute, so it becomes a matter of inclusion of whether a user's statement is "a suitable statement to do a certain question"?
What kind of "user" is this "user"?
There are two problems. The key word is "where?" and whether the entire statement is appropriate to a question like "What happens next?" and whether the entire statement is appropriate for a question like "What happens next?
What kind of "question" is that "question"?
There are two types of templates: one that fits keywords into the template and one that does not take keywords.
What kind of "keywords" are those "keywords"?
democrat
What kind of "demoko" is this "demoko"?
But I wanted to write that keywords can also be multiple words, so the whole sentence should be interpreted as a keyword.
Is there anything else you can tell us about that "keyword"?
In other words, it is good to know the inclusion relationship between keywords, and the inclusion relationship is not a general one, but a binary classification of whether it is appropriate for a particular question
Where are these "keywords" located?
The keywords are extracted from the text and processed, but this is a kind of attention, because the weight of zero or one is multiplied by one for each successive section.
Where are those "key words"?
In a sentence, it's basically one lump. If it's an attention, it could be scattered around, so it's harder to fit it into a template.
What kind of "attention" is that "attention"?
Weighting for a given input sequence
What kind of "mi" is that "mi"?
NGKW Mi
What kind of "weight" is that "weight"?
Values that make the area of interest larger.
The keywords are extracted from the text and processed, but this is just a kind of attention, because the weight of zero or one is multiplied by one for each successive interval.
- Where do these "keywords" come from?
Right now, they're doing conventional keyword extraction and verb-specific extraction.
How do you know what the "key word" is?
https://keicho.netlify.app/#talk=KeOBn7Dr4GJqBUtLnwNS
This page is auto-translated from /nishio/意味は集合ではない、という気づきがあったので、これが何か実用的な応用につながるのか相談したい using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.