Infinite dependency cycles #103
-
|
Hi @grhoten , I am finalizing my proposal, but before I proceed with submission I would love to discuss this case with you.
So, This seems to naturally prevent cycles. Thanks for your time! Looking forward to your thoughts. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
I'm not aware of a scenario where there is an infinite dependency cycle. There is confusion as to what is driving the inflection in a message, which makes it difficult to make a message automatically grammatically correct. You either need a higher level framework to figure that out (a fair amount of NLP), or you need guidance from a translator writing the message. Here is an example phrase. I'm simplifying it by only using English. Assuming that you are replacing only uncle in this single message, the word uncle is both the inflector and inflectee in the message. In the first sentence, it's driving which indefinite article should be put in front of the word uncle. In the second sentence, the word has to be inflected to be genitive (AKA possessive). I can replace uncle with aunt, cousin, brother, sister, and so forth. In a language like French, the thing that is being inserted tends to affect more words than in English. In a language like German, Russian or Finnish, the word being inserted into a message will be more likely to be inflected. Articles and prepositions in one language tends to take on grammatical categories in other languages. So it's definitely hard to extrapolate the rules for one language to another language to determine if the inflection operation is supposed to be an internal or an external inflection operation. Take another example. Do I want to make the object singular to match "is", or do I want to modify "is" to be plural when the object is plural? Both are possible solutions. It's more confusing when you have multiple objects being referenced in a sentence. Take this one as an example. I'm using English for readability. The word "passenger" can be masculine, feminine or unknown (epicene). The "your" can be masculine, feminine or unknown, which is a common problem for pronouncing the possessive pronoun in Arabic, and Hebrew. Should the verb change its grammatical gender to agree with "passenger" or "your"? This disambiguation process is not uniform between languages, and it's not trivial. There are lots of ways to solve this dependency problem, but those solutions are generally outside the scope of Unicode Inflection at this time. This project is primarily reaction based. It's told to make a word singular or plural, and it modifies it as requested. We want to minimize the guessing. I hope that information helps. |
Beta Was this translation helpful? Give feedback.
-
|
Hey George, Thank you for your detailed response. I’ve replied to your message via email earlier yesterday, where I also included my proposal. I believe my proposal addresses many of the inflection aspects, beyond just plurals as I thought this project will cover all the possible grammatical agreements, also mentioned in the provided resources. I have submitted the proposal on the GSoC site so I am now looking forward to your feedback. (The proposal submitted on GSoC website is the final draft) |
Beta Was this translation helpful? Give feedback.
I'm not aware of a scenario where there is an infinite dependency cycle. There is confusion as to what is driving the inflection in a message, which makes it difficult to make a message automatically grammatically correct. You either need a higher level framework to figure that out (a fair amount of NLP), or you need guidance from a translator writing the message.
Here is an example phrase. I'm simplifying it by only using English.
Assuming that you are replacing only uncle in this single message, the word uncle is both the inflector and inflectee in the message. In the first sentence, it's driving which indefinite…