Topic: [BETA] NEW ninjastic.space search (Read 543 times)

Other browsers are equally inaccessible for this tool.
For ninjastics it's still fine

I checked and indeed it was a site issue that was DOWN.
Means it's not a browser problem. Yeah hope this gets resolved soon.

Try another browser.
I experienced same since yesterday but not limited to ninjastics
There are sites I can't assess like coindesk but it works using another browser.
I think this should be a temporary mishap and solutions would be available soon.

I always search for something in the forum always use this tool, because it is easy and simple but now this site cannot be accessed is there an update or is it interference?

@TryNinja might be able to clarify? Because it shows “Page not working”.

I do like how it works. A lot! I see it as a leap forward in both the way the query can be done and the way the results are shown. Great work!

One small thing is missing imho: I think that ninjastic.space main page should have somewhere a link to the beta search page.

I did some basic testing and it is loading results very fast, but I would prefer if you could add dark theme switch.
It would also be great if you could add separate fields for author, boards and key words, or some other type of separation with suggestions.
Posting simple instructions on the bottom of the page would be helpful for newbies, as well as any other simplification.

Ok, figured out.

It's related to this:

"Overfeat" comes right after an URL, which becomes messed up when they are right after/before a line break.

So this:

Becomes this, in my index:

Instead of the token/word Overfeat, it is saved as an entire url: "https://bitcointalksearch.org/topic/m.11414986Overfeat"

I've fixed this already but I haven't reindexed the posts because there is no need to do that right now (waiting because if I find another issue, then I need to reindex a third time).

Thanks. I messed up my search query: it's quick and dirty on the command line, and still had an "img" search in the string. Without it, I find 3 matching posts:

I could find the post shield132 was looking for using forum search engine.
shield132 mentioned the word "Overfeat" in this post. The post has not been edited and your archive and ninjastic.space both have that post.

I did a (local) search on my posts archive, and can't find it either. If you added it to a post by editing it, I won't have it.

I think you should write down search examples on your website to make it clear for people how to search. That placeholder is not good because once I type anything, it disappears and I need to remember it or open another tab with an empty form to read the placeholder and edit it according to my requirements.

In overall, it loads quickly and nicely but I think that the search has some problems. For example, I tried to search author:shield132 "Overfeat" but it didn't load anything while I have mentioned the word Overfeat in my posts before but ninjastic.space's search can't find it.

Search speed is impressive and my first short test works fine. Now I know I've quoted theymos 659 times
One suggestion: the small sample in gray on the site ("author: satoshi, ............") is gone when I start typing. It would be nice to show all possible search tags under this field.

Must reach those post quotas! I've hit Ignore on him now.

Unless you're the click-only-generation, then typing is very confusing

---

---

I'll test it more when I actually have something I'm trying to find back.

The complexity involves getting the algorithm to automatically define what is or is not spam.


Maybe created a new tag ex_board_id, which allows you to exclude a certain board from the search.

The objective here is obviously not to "destroy" Google, but to make a search system for Bitcointalk that is more useful than Google's site:bitcointalk.org (and the native search).

So yes, I know it's going to be hard, but I don't see it being that complex. For one thing, we have only millions of posts, not billions of pages like google. Second, we won't have as many users, and consequentially people trying to "abuse" the system compared to Google (only a few thousand people, instead of millions of people using the search).

Regarding spam, I don't think anyone's going to shed any tears over spam posts not being indexed. For instance, I plan to (and am currently in the process of) index the entire Politics and Society. However, you can eliminate 95% of the spam by enforcing a minimum character count in posts like in sig campaigns, as well as using a regex to filter out sig/bounty application posts. Google's bias accusations are literally because they tweak the ranking of certain kinds of pages that they are offended by - I have no such plans.

Now yes, I know that if you're trying to conjuncture your own search software, you will run into the limitations you just described.

However, we are very fortunate to have open-source search software written by people who do have years of experience in the field, for example Apache Lucene (which coincidentally I have experience with but in a completely different industry). Because of this, sites can roll up their own search, in case Google's results are insufficient. Example: Hacker News.

The above is not written to discourage anybody from creating their own search algorithms, it just shows the requirements that need to be in place to make an "intelligent" search. There's nothing wrong with the current implementation if you're just using it to search for someone's (or your own) post, and I think that is what TryNinja is trying to make here.

I can see much value there will be for an intelligent search engine if somebody manages to pull it off, especially with other repositories like bitcoin-dev (the old version) in dire need of indexing as well.

Do you have any idea how many years it took for search engines to do what you're asking for?
A system of this kind is very complex to build. That's why it remains difficult to destroy Google and its algorithm.


What TryNinja could do is a classification, where users who used the search classified the result as positive (close to what was intended) or negative (spam). And the system began to place the results with the most negative grades at the end of the list. Even so, there would be a danger of making the information biased.

Nice job Ninja; you are obviously the leader in post searching, and I for one would not want to compete with you. 

You have a chance here to implement AI into your search; a third party service could identify what information the user is looking for, instead of what keywords the user is looking for. 

Well, good luck on your journey on recreating Google for BitcoinTalk. You're going to need it.

I don't know how nor want/have the the time to do all of that, so I'm ok with facilitating searching on my archive (which was always a subproject of my telegram bot) the way I'm doing it. I've used it myself many times to find quotes and references, so I can tell you that it's not an useless tool.

What you want is for someone to search "how to create a electrum wallet" and find out all the main OPs and pieces of text which correctly teach how to do that. Without doing AI'ish stuff and/or manually tagging posts I can't see how that would work. But I honestly hope you can make it.

Not technically viable right now.

Is it possible to include a tag on a post that already deleted on search result? Sometimes I  find a deleted post which I’m about to reply but it turns out to be deleted or edited already.

I’m not sure if this is possible but this will be awesome at least for me that use search function to find a topic I need instead of the forum search feature.

This is going to sound controversial but I think that meaningless posts should not be indexed, or at least not returned. The algorithm should be prioritizing information over exact matches.

Open DuckDuckGo or any other search engine and search for it. After the videos, images etc you will start seeing results like this:



The mission of a search engine is to help people find what they are looking for. This is almost always stuff like answers, facts, and other kinds information with a semantic load. There is a lot of spam on Bitcointalk that nobody wants to read on the main site. So as much as possible, search engines must avoid returning spammy posts as results because this will discourage people from using them, as they will not find the kind of information they are looking for.

PS. even merit cannot be used to bias the posts because author merit does not apply if the post in question has no semantic load, and post merit allows for merit farms to manipulate search results by meriting spammy posts.

And what do you expect it to find when you search "I like trains"? Or any actual example, for that matter.