0
$\begingroup$

I was told that my question was missing some specific details, which I fixed. A day later my question was closed due to missing details and clarity. What details?

I fixed the problems that were highlighted in the comments and it was closed with no further explanation.

Here is the link to the post: Training an agent to choose a string from a list of strings

$\endgroup$

1 Answer 1

3
$\begingroup$

Closing a question is (usually) not a unilateral decision; it is usually only closed when multiple people have voted for it to be closed. An exception is that moderators can directly close, but that is not what happened here; three different people voted to close.

It is of course possible that they each of slightly different reasons for voting to close / different details they think are missing. I can't speak for them. I have taken a look though and can explain what I would recommend taking a look at to improve the clarity:

The title is not descriptive, it does not summarise the core of the question: "Training an agent to choose a string from a list of strings".

That title somewhat sets the context of the question (it's about training an agent in an environment where it needs to choose strings from some list of candidates apparently), but not the core of the question. To me, the core of the question seems to be that you're curious about how the actions should be represented; should you represent them simply by giving every possible word in the vocabulary a unique index, or should you actually use a multi-dimensional representation where you represent actions as the arrays of characters / strings.

Furthermore, directly just mentioning the name of the game you're interested in (Wordle) is probably more descriptive than the somewhat vague description of the task as "to choose a string from a list of strings". Lots of other tasks could be described like that too. So, I think I would recommend a title more like:

How to encode actions for training a Wordle agent?


Apart from that, I think the main body of the question itself right now is reasonably clear. It is possible that some of the people who voted to close actually already did before you edited, or maybe they still see things they find unclear.

One thing that could maybe help would be to more explicitly state whether you really really want to use a Reinforcement Learning solution (for instance, because you want to learn more about that specific field), or whether you're just looking for any AI solution in general. Otherwise, you get comments like Neil's first comment, for example, when people may believe that RL may not necessarily be the best/easiest solution, but are not sure whether or not that's important for your question.

$\endgroup$
1
  • 2
    $\begingroup$ Thanks for the feedback. I edited my original question accordingly. $\endgroup$
    – Yuval Amir
    Commented Feb 13, 2022 at 12:47

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .