For a simple test, I searched "fall of the roman empire". In your search engine, I got wikipedia, followed by academic talks, chapters of books, and long-form blogs. All extremely useful resources.
When I search on google, I get wikipedia, followed by a listicle "8 Reasons Why Rome Fell", then the imdb page for a movie by the same name, and then two Amazon book links, which are totally useless.
The real trick would be some kind of engine that can aim just above where the user's at.
But no, these stories all come from cookie-cutter "new media" blog sites, written by an anonymous content writer who's repackaged Wikipedia/Discogs info into Buzzfeed-style copy writing designed to get people to "share to Twitter/FB". No passion, no expertise. Just eyeballs at any cost.
It reminds me of an annoyance I have with the Kindle store. If I wanted to find a book on, let's say, Psychology, there is no option to find all-time respected books of the past centenary. Amazon's algorithms constantly push to recommend the latest hot book of the year. But I don't want that. A year is not enough time to have society determine if the material withstands time. I want something that has stood the test of time and is recommended by reputable institutions.
They do? That would explain a lot - but ironically, I can't find a good source on this. Do you have one at hand?
I think they take different things into account based on the thing being searched.
Engagement, as measured in clicks and time spent on page, plays a big part.
But you're right, to a degree, as frequently updated pages can rank higher in many areas. A newly published page has been recently updated.
A lot depends on the (algorithmically perceived) topic too. Where news is concerned, you're completely right, algos are always going to favor newer content unless your search terms specify otherwise.
PageRank, in it's original form, is long dead. Inbound link related signals are much more complex and contextual now, and other types of signals get more weight.
The only definitive source on this would be the gatekeeper itself. But Google never says anything explicitly, because they don't want people gaming search rankings. Even though it happens anyway.
It's all stamped with Google Ads, of course, and then Google ranks these pages high enough to rake in eyeballs and ad dollars.
Also there's the fact that each year, the average webpage picks up two more video elements / ad players, one or two more ad overlays, a cookie banner, and half a dozen banner/interstitials. It's 3-5% content spread thinly over an ad engine.
The Google web is about squeezing ads down your throat.
The only reason people make content they aren't passionate about is advertising.
The problem is that clickbait and low effort articles could be good enough to get the click, but low effort enough to drag society into the gutter. As time passes, the system is gamified more and more where the least effort for the most clicks is optimized.
Actually that's not always the case. We publish a lot of blog content and it's really hard to publish new content that replaces old articles. We still see articles from 2017 coming up as more popular than newer, better treatments of the same subject. If somebody knows the SEO magic to get around this I'm all ears.
1 Accept quid pro quo to send all queries to Google by default
If what these companies were telling their readers was true, i.e., that advertising is "essential" for the web to survive, then how are the sites returned by this search engine for text-heavy websites (that are not discoverable through Google, the default search engine for Chrome, Firefox, etc.) able to remain online. Advertising is essential for the "tech" company middleman business to survive.
A "Top 10 albums of all time" post is actually better off going through 10 genres of popular music from the past 50 years and picking the top album (plus mentioning some other top albums in the genre) for each one.
That gives the user the overview they're probably looking for, whether those are the top 10 albums of all time or not. It's a case of what the user searched for vs what they actually really want.
So, AIs are actually on par with most adults now? (Sorry)
I'm not sure the idea of a sentient being not having a bias is meaningful. Reality, once you get past the trivial bits, is subjective.
I want an AI that is biased to the truth when there is an objective one, and my tastes otherwise. (that is when asked to find a good book it should give me fantasy even though romance is the most popular genre and so will have better reviews)
Maybe not the same kind of bias we think of in terms of politics and such, but I wonder if there's a connection.
You have to e causality reversed. Google results reflect the fact that society is dumb.
Appropriately enough, I couldn't find a good quote to verify that since Google is only giving me newspapers and magazines talking about Sir Tim in the context of current events. I do believe it's in his book "Weaving the Web" though.
I hear stories about Flash and ActiveX but I literally never needed these to shop or pay bills online. Payments also didn't require scripts from a dozen domains and four redirects..
> If you are looking for fact, this is almost certainly the wrong tool. If you are looking for serendipity, you're on the right track. When was the last time you just stumbled onto something interesting, by the way?
So if you for example compare "SDL tutorial" with "SDL tutorials". On google you'd get the same stuff, this search engine, for better or worse doesn't.
This is a design decision, for now anyway, mostly because I'm incredibly annoyed when algorithms are second-guessing me. On the other hand, it does mean you sometimes have to try different searches to get relevant results.
On the other hand, Google’s unawareness of (extensive and ubiquitous) Russian noun morphology is essentially what allowed Yandex to exist: both 2011 Yandex and 2021 Google are much more helpful for Russian than 2011 Google. I suspect (but have not checked) that the engine under discussion is utterly unusable for it. English (along with other Germanic and Romance languages to a lesser extent) is quite unusual in being meaningfully searchable without any understanding of morphology, globally speaking.
I... don’t think anything particularly surprising is happening here, except for quotes being apparently ignored? I’ve had it explained to me that a rare word is essentially indistinguishable from a popular misspelling by NLP techniques as they currently exist, except by feeding the machine a massive dictionary (and perhaps not even then). BRST is a thing that you essentially can’t even define satisfactorily without at the very least four years of university-level physics (going by the conventional broad approach—the most direct possible road can of course be shorter if not necessarily more illuminating). “Best” is a very popular word both generally and in searches, and the R key is next to E on a Latin keyboard. If you are a perfect probabilistic reasoner with only these facts for context (and especially if you ignore case), I can very well believe that your best possible course of action is to assume a typo.
How to permit overriding that decision (and indeed how to recognize you’ve actually made one worth worrying about without massive human input—e.g. Russian adjectives can have more than 20 distinct forms, can be made up on the spot by following productive word-formation processes, and you don’t want to learn all of the world’s languages!) is simply a very difficult problem for what is probably a marginal benefit in the grand scheme of things.
I just dislike hitting these margins so much.
(DDG docs do say it supports +... and even +"...", but I can’t seem to get them to do what I want.)
There are six of them (nominal [subject], genitive [belonging, part, absence, “of”], dative [indirect object, recipient, “to”], accusative [direct object], instrumental [device, means, “by”], prepositional [what the hell even is this]), so you have (cases) × (numbers) = 6 × 2 = 12 noun forms, and adjectives agree in number and gender with their noun, but (unlike Romance languages) plurals don’t have gender, so you have (cases) × (numbers and genders) = 6 × (3 + 1) = 24 adjective forms.
None of this would be particularly problematic, except these forms work like French or Spanish verbs: they are synthetic (case, number and gender are all a single fused ending, not orthogonal ones) and highly convoluted with a lot of irregularities. And nouns and adjectives are usually more important for a web search than verbs.
https://e.humanities.uva.nl/publications/2004/kamp_lang04.pd...
Re compounds, I expected they would be more or less easy to deal with by relatively dumb splitting, similar to greedy solutions to the “no spaces” problem of Chinese and Japanese, and your link seems to bear that out. But yeah, cheers to more language-specific stuff in your indexing. /s
Hmm I seem to be getting only relevant results, no "best", not sure what you mean. Are you not doing verbatim search?
Let's name it after a famous old scientist, and maybe add the year to prove it's modern: Galileo 2021.
There was such an app in the early 2000's, before Google went mainstream, and Altavista-like engines were not good: Copernic 2000.
I guess I'm officially old now.
This is a pattern I see over and over again, some research group or academics show that something can be done (summaries that make sense and are true summaries, evolutionary algorithm FPGA programming, real time gaze prediction, etc) and there's a few published code repos and a bit of news, then 'poof' - no where to be seen for 15 years or more.
https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theore...
You can't combine a few different ranked lists and expect to get results better than any of the original ranked lists.
I am skeptical of this application of the theorem. Here is my proposal:
Take the top 10 Google and Bing results. If the top result from Bing is in the top 10 from Google, display Google results. If the top result from Bing is not in the top 10 from Google, place it at the 10th position. You'd have an algorithm that ties with Google, say 98% of the time, beats it say, 1.2% of the time, and loses .8% of the time.
https://ccc.inaoep.mx/~villasen/bib/AN%20OVERVIEW%20OF%20EVA...
and the first conclusion is "something that you think will improve relevance probably won't"; the TREC conference went for about five years before making the first real discovery
https://en.wikipedia.org/wiki/Okapi_BM25
It's true that Arrow's Theorem doesn't strictly apply, but thinking about it makes it clear that the aggregation problem is ill-defined and tricky. (e.g. note also that a ranking function for full text search might have a range of 0-1 but is not a meaningful number, like a probability estimate that a document is relevant, but it just means that a result with a higher score is likely to be more relevant than one with a lower score.)
Another way to think about it is that for any given feature architecture (say "bag of words") there is an (unknown) ideal ranking function.
You might think that a real ranking function is the ideal ranking function plus an error and that averaging several ranking functions would keep the contribution of the ideal ranking function and the errors would average out, but actually the errors are correlated.
In the case of BM25 for instance, it turns out you have to carefully tune between the biases of "long documents get more hits because they have more words in them" and "short documents rank higher because the document vectors are spiky like the query the vectors". Until BM25 there wasn't a function that could be tuned up properly and just averaging several bad functions doesn't solve the real problem.
Conditions the mixed ranker doesn't have to satisfy "ranking while also meeting a specified set of criteria: unrestricted domain, non-dictatorship, Pareto efficiency, and independence of irrelevant alternatives"
Hypothetically you could treat these functions as meaningful but when you try you find that they aren't very meaningful.
For instance IBM Watson aggregated multiple search sources by converting all the relevance scores to "the probability that this result is relevant".
A conventional search engine will do horribly in that respect, you can fit a logit curve to make a probability estimator and you might get p=0.7 at the most and very rarely get that, in fact, you rarely get p>0.5.
If you are combining search results from search engines that use similar approaches you know those p's are not independent so you can't take a large numbers of p=0.7's and turn that into a higher p.
If you are using search engines that use radically different matching strategies (say they return only p=0.99 results with low recall) the Watson approach works, but you need a big team to develop a long tail of matching strategies.
If you had a good p-estimator for search you could do all sorts of things that normal search engines do poorly, such as "get an email when a p>0.5 document is added to the collection."
For now alerting features are either absent or useless and most people have no idea why.
Suppose there's an unambiguous ranked preference by all people among a set (webpages, ranking). Suppose one search engine ranks correctly the top 5 results and incorrectly the next 5 results, while another ranks incorrectly the top 5 and correctly the next 5.
What can happen is that some there may be no universally preferred search engine (likely). In practice, as another commenter noted, you can also have most users prefer more a certain combination of results (that's not difficult to imagine, for example by combining top independent results from different engines for example).
Still, the best way to break SEO is to have actual competition in the search space. As long as SEO remains focused on Google there is an opportunity for these companies to thrive by evading SEO braindamage.
Like an article about some current event will undoubtedly begin with "when I was traveling ten years ago...".
You see some of that "two readers" divide sometimes even in classic cookbooks, where "celebrity" chefs of the day might spend much of a cookbook on a long rambling memoir. Admittedly such books were generally well indexed and had table of contents to jump right to the recipes or particular recipes, but the concept of "long personal ramble of what these recipes mean to me" is an old one in cookbooks too.
A couple of browser add-ons specifically geared around trimming recipe pages down have been taken down due to similar complaints.
One audience matches your description and is the invested reader. They want that blogger's story telling. they might make the recipe, but they're a dedicated reader.
The other audience is not the recipe-searcher, but instead Google. Food bloggers know that recipe-searchers are there to drop in, get an ingredient list, and move on. They won't even remember the blog's name. So the site isn't optimized for them. It's optimized for Google.
"Slow the parasitic recipe-searcher down. They're leeches, here for a freebie. Well they'll pay me in Google Rank time blocks."
This is not entirely true, though. If a randomly found recipe turns out particularly good, I'll bookmark the site and try out other dishes. It's a very practical method to find particularly good* recipe collections.
*) In this case "good" means what you need - not just subjectively "tasty", but e.g. low cost, quick to prepare, low calorie or in line with a particular diet and so on.
I think this assumes facts not in evidence. It certainly seems like an overwhelming number of "blogs" are not actual blogs but SEO content farms. There's no regular readers of such things because there's no actual authors, just someone that took a job on Fivver to spew out some SEO garbage. Old content gets reposted almost verbatim because new results better according to Google.
The only reason these "blogs" exist is to show ads and hopefully get someone's e-mail (and implied consent) for a marke....newsletter.
It's at least half the business models of Food Network shows: aspirational kitchens and the people that live in them and also sometimes here's their recipes. (The other half being competitions, obviously.) I've got friends that could deliver entire doctoral theses on the Bon Appetit Test Kitchen (and its many YouTube shows and blogs) and the huge soap operatic drama of 2020's events where the entire brand milkshake ducked itself; falling into people's hearts as "feel good" entertainment early in 2020/the pandemic and then exploding very dramatically with revelations and betrayals that Fall.
Which isn't to say that there aren't garbage SEO farms out there in the food blogging space as well, but a lot of the big ones people commonly complain about seeing in google's results do have regular fans/audiences. (ETA: And many of the smaller blogs want to have regular fans/audiences. It's an active influencer/"content creator" space with relatively low barrier to entry that people love. Everyone's family loves food, it's a part of the human condition.)
That's ads. When mobile users have to scroll past 10 add, theyll click on some of them and make the blog money.
I continue to be curious about this kind of complaint. If all you want is a recipe list, without any of the fluff, why would you click on a link to a blog, rather than on a link to a recipe aggregator?
Foodie blogs exist specifically for the people who want a foodie discussion and not just an ingredients' list.
Is it because blogs tend to have better recipes overall? In that case, isn't there a bit of entitlement involved in asking that the author self-sacrificingly provides only the information that you want, without taking care of their own needs and wants, also?
If you want JUST recipes, pay money instead of just randomly googling around. America's test kitchen has a billion, vetted, and really good recipes. That solves that problem.
To be honest, I don't follow recipes when I cook unless it's a dish I've never had before. At that point what I want is to understand the point of the dish. A list of ingredients and preparation instructions don't tell me what it's supposed to taste and smell like. The foodie blogs at least try to create a certain... feeling of place, I suppose, some kind of impression that guides you when you cook. I wouldn't say it always works but I appreciate the effort.
My real complaint with recipe writers is that they know how to cook one or two dishes well and they crib the rest off each other so even with all the information they provide, you still can't reliably cook a good meal from a recipe unless you've had the dish before. But that's my personal opinion.
> “Mere listings of ingredients as in recipes, formulas, compounds, or prescriptions are not subject to copyright protection. However, when a recipe or formula is accompanied by substantial literary expression in the form of an explanation or directions, or when there is a combination of recipes, as in a cookbook, there may be a basis for copyright protection.”
I also think a search engine like this would be quite hard to game. An ML-based classifier trained on thousands of text-heavy and media-heavy screenshots should be quite robust and I think would be very hard to evade, so the "game" will become more about how identify the crawler so you can serve it a high-ranking page while serving crap to the real users, and it seems fairly easily to defeat if the search engine does a second pass using residential proxies and standard browser user agents to detect this behavior (it could also threaten huge penalties like the entire domain being banned for a month to even deter attempts at this).
So just add human review to the mix, if a site is obviously trying to game the system (listicles, seo spam etc) just drop and ban them from the search index.
If you almost only plant wheat, you are going to end up with one hell of a pest problem.
If you almost only have Windows XP, you are going to have one hell of a virus problem.
If you almost only have SearchRank-style search engines (or just the one), you are going to have one hell of a content spam problem.
Even though they have some pretty dodgy incentives, I don't think google suffers quality problems because they are evil, I think ultimately they suffer because they're so dominant. Whatever they do, the spammers adapt almost instantly.
A diverse ecosystem on the other hand limits the viability of specialization by its very nature. If one actor is attacked, it shrinks and that reduces the opportunity for attacking it.
but arent you curious about the 7th reason? it will surprise you!
If you move data organization to another type of organization (non-profit, state, universities - private or public), then the question of data prioritization becomes highly political. What should be exposed? What should not? What to put first? ...
It is already, but to a smaller extend since money-making companies have little interest in data meaning, and high interest in the commercial value of their users.
Google's revenue last year was 146 billion, and it operates not anywhere near the theoretical maximum. Most of that revenue is advertisement.
Here's what I've tried with a few variations: golang generics proposal, machine learning transformer, covid hospitalization germany
[edit] formatting
Another need I guess might be reviews, for which RT or MC are better than IMDB: not sure if either of those two will fare better than IMDB in this search engine but again Wiki has links out (in addition to good reception summaries)
I never even posted on it myself, but browsing the discussions one could learn all sorts of trivia, inside info, speculation, etc about each movie.
Since they (inexplicably) killed that feature, I rarely even visit anymore. Your right, for many purposes wikipedia is better, especially for TV series episode lists with summaries.
I think they removed it in part because new movies, like star wars and superhero movies, had alot of negative activity.
https://search.marginalia.nu/search?query=harry+potter+fanfi...
This seems to return a pretty decent number of sites relating to that (as well as some sites not relating to that).
The search engine isn't always great at knowing what a page is about, unfortunately.
This seemed to return mostly relevant results
https://search.marginalia.nu/search?query=%22harry+potter%22...
https://search.marginalia.nu/search?query=Best+%22harry+pott...
This used to be how all search engines worked, but I guess people have been taught by google that they should ask questions now, instead of search for terms.
I wonder how I can guide people to make more suitable queries. Maybe I should just make it look less like google.
What kind of links where you expecting to find?
If your goal was to create a search engine that ignored listicles and other fluff and instead got you meatier results like "academic talks" and such, then no.
First result after Wikipedia:
"Radiophone Transmitter on the U.S.S. George Washington (1920)
In 1906, Reginald Fessenden contracted with General Electric to build the first alternator transmitter. G.E. continued to perfect alternator transmitter design, and at the time of this report, the Navy was operating one of G.E.'s 200 kilowatt alternators http://earlyradiohistory.us/1919wsh.htm "
Another result in the first few:
" - VANDERBILT, GEORGE WASHINGTON
PH: (800) ###-#233 FX: (#03) 641-5###. https://www.ScottWinslow.com/manufacturer/VANDERBILT_GEORGE_... "
And just below that terrible result:
"I Looked and I Listened -- George Washington Hill extract (1954)
Although the events described in this account are undated, they appear to have occurred in late 1928. I Looked and I Listened, Ben Gross, 1954, pages 104-105: Programs such as these called for the expenditure of larger sums than NBC had anticipated. It be http://earlyradiohistory.us/1954ayl2.htm "
Dramatically worse than Google.
---
Ok, how about a search for "Rome" then? Surely it'll pull some great text results for the city or the ancient empire.
First result after Wikipedia:
"Home | Rome Daily Sentinel
Reliable Community News for Oneida, Madison and Lewis County http://romesentinel.com/"
The fourth result for searching "Rome":
"Glenn's Pens - Stores of Note
Glenn's Pens, web site about pens, inks, stores, companies - the pleasure of owning and using a pen of choice. Direcdtory of pen stores in Europe. http://www.marcuslink.com/pens/storesofnote/roma.html"
Again, dramatically worse than Google.
---
Ok, how about if I search for "British"?
First result after Wikipedia:
"BRITISH MINING DATABASE
British_Mining_Database http://www.users.globalnet.co.uk/~lizcolin/bmd.htm "
And after that:
"British Virgin Islands
Many of these photos were taken on board the Spirit of Massachusetts. The sailing trip was organized by Toto Tours. Images Copyright © Lowell Greenberg Home Up Spring Quail Gardens Forest Home Lake Hodges Cape Falcon Cape Lookout, Oregon Wahkeena http://www.earthrenewal.org/british_virgin_islands2.htm"
Again, far off the mark and dramatically worse than Google.
I like the idea of Google having lots of search competition, this isn't there yet (and I wouldn't expect it to be). I don't think overhyping its results does it any favors.
As for the results you linked, it's part of the zeitgeist to list other entities sharing the same name. Sure, they could use some subtle changes in ranking, but overall the returned links satisfy my curiosity.
> If you are looking for fact, this is almost certainly the wrong tool. If you are looking for serendipity, you're on the right track. When was the last time you just stumbled onto something interesting, by the way?
If you are going top claim something is wide of the mark then you really ought to tell us at least roughly where the mark is.