In the search of a question

« first day (1156 days earlier) ← previous day next day → last day (1610 days later) »

user12692

12:55 AM

There has be a conversation about searching in the CRUDE chat room:

user12692

@BillDubuque Before the linked dupe is edited to be located more easily, your deletion is making this site rather inconvenient. If someone is going to need to ask the same question about $x^2+35=7^n$ tomorrow, I have no idea how he or she would need to search in order to identify the abstract dupe you suggest.

user12692

in CRUDE, 3 hours ago, by Bill Dubuque

D1 (bis) this dupe FAQ needs 2 more votes (again, sigh)

user12692

I am wondering if there is any possibility that Approach0 could identify the claimed duplicate in this case.

Martin Sleziak

2:20 AM

Somewhat related to the above and to this:

in CRUDE, 1 min ago, by Bill Dubuque

In any case, if you think those particular parameters will be searched for (unlikely) then you could simply add them to the body of abstract dupe question (e.g. "some other common instances are ....").

Question. Are posts reindexed to Approach0 corpus (if that's the correct term) after they have been edited?

For example, if the equation $x^2+3=3^n$ is added to some post (which did not contain it originally), will the post appear among the search results eventually?

Martin Sleziak

2:33 AM

Sorry, I meant to link to the search for $x^2+3=3^n, which corresponds to the formula which I mentioned as an example.

2 hours later…

Wei Zhong

4:56 AM

@Sil What is the most recent commit of the code you use to generate that searchd?

Wei Zhong

5:08 AM

@Sil also, are you using the same code that you generate the index to build searchd? And when you compile indri, ensure you follow this approach0.xyz/docs/src/…

@Sil Alternatively, clone my forked version of Indri. That will make your build configuration the same as mine, and if problem persists, I will use ubuntu and try to reproduce this issue.

2 hours later…

Wei Zhong

7:10 AM

@Sil I think I have reproduced the overflow issue on a ubuntu distro. None of my above guess was right, so let me fix it and get back to you soon.

2 hours later…

Wei Zhong

8:47 AM

@Sil Try pull the newest code and see if it has fixed the segmentation fault, particularly this commit: github.com/t-k-/a0-private/commit/…

There is another segmentation fault only when searchd existing on Ubuntu environment, it seems the MPI_Finalize function has caused this problem, but I have not figured out the reason. If you ignore this, searchd on Ubuntu works fine.

@MartinSleziak No, only after corpus is fetched again from MSE website, and re-indexing the updated corpus files.

Martin Sleziak

So we cannot rely on the fact that Approach0 will contain the edited version of a post, if the edit was made after the post was indexed. Did I understand that correctly?

Wei Zhong

So if the index (built offline) has not been updated, the search engine has no way to find the new information.

Right, after indexing, search engine does not fetch anything from MSE or other website at all.

Martin Sleziak

9:02 AM

Thanks for the response!

Wei Zhong

It is not like the search engine goes and visits the (indexed) link and search content like a web Browser. That is too slow.

No problem, it is good to make this matter clear.

Sil

9:26 AM

@WeiZhong sure will check later today, thanks for the quick fix

1 hour later…

Wei Zhong

10:44 AM

I believe the issue of the later segmentation fault is an environment issue. After compiling a simple program that invokes both MPI and libevent functions on Ubuntu, I reproduced the segmentation fault.

Wei Zhong

11:07 AM

Posted a question about this on StackOverflow: stackoverflow.com/questions/59137563

2 hours later…

Wei Zhong

12:57 PM

@Sil With regarding to aops350187, actually you can find it using query: Functional, $f(x^3+y^3)=xf(x^2)+yf(y^2)\ ,\ (\forall)x,y\in \mathbb{R}$

here is the link

But if you only search terms like "Functional" or "Marathon", we hardly find it in search results because the way text query scoring works in search engine. The similarity scoring will penalize over-length document.

I setup a local experiment with only that "topic" file indexed, and if you search "Functional" and "Marathon" you will get only a score of -5.351.

4 hours later…

Sil

4:41 PM

@WeiZhong I can confirm that query works, although the original functional equation f(f(x)y+x)=f(x)f(y)+f(x) is not found, see 45.79.114.212/search/…

5 hours later…

Sil

10:03 PM

Also the fix of segmentation fault works, now I was able to make complete end to end scenario work, having corpus indexed and searching on my local machine!

« first day (1156 days earlier) ← previous day next day → last day (1610 days later) »

all rooms