I think what's happening here is that there's a very impressive and sophisticated heuristic for predicting the probability of what you want to type by looking at the frequency of what you have typed in similar contexts. It uses its state-of-the-art AI to evaluate the context and build an array of candidate words, along with their respective probabilities. I suspect it is very accurate as it does this. Then it sorts the array by probability and pops the top element into the predictive text input.
Alas -- per my pet theory anyways -- it sorts like this:
candidateWords.sort((a,b) => a.probability - b.probability);
candidateWords.sort((a,b) => b.probability - a.probability);
And yet, when the latter inevitably breaks on an edge case, users can try to fix it themselves. They don't hit a wall of frustrated "I can't do anything", they hit a challenge that they are empowered to try overcoming. They already know what they want and can set things that way, rather than trying vaguely to teach a system (machine learning, hardcoded heuristics engine, department of humans making seemingly unconnected changes to a GUI with each passing version and no obvious plan) to understand their desires.
I miss the days when users were seen as intelligent professionals who are willing to change settings, create and re-dock an assortment of toolbars to every edge of every screen/window to suit their daily tasks, read a manual (or at least search the integrated help entries) to overcome problems. Rather than "busy" phone users who just want to complete a task with minimal time spent learning and get back to posting on facebook or whatever, and who accept the automagical solution because adequate results instantly are somehow considered better than great results with some work.
Ugh, that whole block of text just kept growing; I had better leave and go ramble/rant at trees or clouds or something elsewhere.
This works fine for me with GBoard. Are you drawing a little circle on the o to indicate you want the double letter?
The weird thing is that if I type nothing at all, the contextual predicted "next word" on GBoard is actually very good -- I wasn't praising it for comedic effect. But it really does seem like there's a sign error in a sort function which kicks in after you start typing.
I've specifically googled for instructions on how I might be expected to use the swipe-style keyboards, and turned up nothing.
edit: Don't actually see a tutorial in the app. Maybe I'm confused with another app such as Swype, but the same technique seems to apply to all.
About it only replacing after you start writing, the probability of "son" must grow faster than the one of "soon" after you type "so". If there were such a huge bug, you would be seeing a new word every time, not always the same one.
All my devices insist on shaming me (or autocorrecting me) for one of those spellings. At this point it feels like a complete gamble which I'll get corrected on. I'm just slowly getting used to correcting the autocorrect. :(
The only thing it is confused about is the letter "i", which means "and" in Bosnian, and SwiftKey often capitalizes it where it shouldn't. Probably happens because I often mix Bosnian and English in the same message (instead of translating technical terms).
I'd love a feature to disable all that deep learning and AI and just use the algorithm they originally had (proximity of where you typed to words in the dictionary). That worked so much better.
I had a Galaxy S3 and was a heavy user of Swype. My friends marveled at how fast I could type with it. It was perfect! I recently changed my phone to an S8 and Swype became unusable. It gets almost every second word wrong, so much that I'm thinking of disabling it entirely :(
iOS keyboards have been getting worse with every update, probably as more and more engineers feel the need to make a mark, or are required to fix bugs, and they spoil it. The simple and predictable statistical model of the original iOS was better than what we have now. So much of iOS was better back then, IMHO.
Honestly it's very probable that I'll only ever keep a smart phone around as a music/podcast/audiobook device.
As the saying goes, smart phones are just pocket computers with shitty phones attached.
Once you hit space and start a new word it’s game over.
Frequently happens at the worst moment when you’re trying to mash out a complex explanation in a rushed fury.
When wages don’t grow over 10 years, what incentive is there to write the best software you can?
The majority of people don't give a shit. Correct spelling and obscure searches are not even on their radar, it's not a part of their reality. Don't let the comments here fool you -- it is a very specific, picky, technical crowd that frequents HN.
The voice of "those who care" has always been a minority, although it used to matter more, simply because people who care and worry and try to do a good job tend to have more power and money (conscientiousness is a great predictor of success), and so businesses cater to them more. Now that everything seems to be turning more uniform, more global, more binary, more equal, that voice is marginalized (good thing? bad thing?) -- you're seeing the effect of a hoi polloi stampede.
So it's not the fault of "incompetent programmers" -- it may be a trickle down effect of our social incentives and economic trade-offs.
To add to this, people who cares most powerfully had probably switched to alternative, open source, software solutions. This leaves the remaining group with less "care" on average so fewer would complain. Kinda like evaporative cooling.
I think another dimension is how deployability has changed.
Before... When you wrote and shipped software, getting your software out was a big problem, a big deal. This also meant that if you shipped a bug, shipping an update would be equally expensive (for you and your customers), and the amount of goodwill you lost would be quite tremendous.
Now everyone has a appstore, always up to date apps, and whatever else is usually "in the cloud" somewhere. The time of people installing applications in a normal desktop-context, with installers and having IT-administrators handle updates once every second year is surely long gone.
With that kind of change, and an increased focus on delivering early, doing proper QA is no longer something which is rewarded in the market.
Who cares if you made a bug-free, awesome service, when you did it 6 months after someone else shipped a similar, but buggy service which everyone is already using? They have established a user-base and as such already has social momentum and lock-in.
What do you have to offer which is not only fantastic enough to make some bother migrating, but also so amazing that these people will also go convert their friends and families? "Less bugs" alone is not going to cut it.
Basically, taking the time to deliver quality software these days is increasingly something you get punished for in the market-place.
The result? We get shit like this and we can only blame ourselves.
If enough people start suing or asking back for their money, companies will surely improve their QA.
I use Minuum - looks whacko at first but let's me use way more of the screen when replying and it is really accurate!
Unfortunately I can't remember any examples at the moment, it's just something that happens to me every so often. They're really irritating though, because they aren't well expressed in the current autocorrect UI (which works on the current word) and it doesn't seem to get the hint when you go back and correct it, so it keeps applying it over and over.
The corpora used for popular word embeddings are full of weird nonsense text (in the case of word2vec) or autogenerated awfulness like spam and the text of porn sites (in the case of fastText) or both (GloVe). And most people who implement ML don't care how their data is collected.
Or don't use a machine learning model. I honestly don't care, just don't automatically turn a correct "its" into an incorrect "it's".
Do publishers sell their published text as a mass for use in AI/ML? Like 1000 books, no images or frontispiece, etc., possibly jumbled by sentence/para/page.
They say they welcome contributions; I don't know if they just mean new sources of text, or if this includes code for filtering or fixing their existing ones.
How to I try the AOSP keyboard? I have a Google Pixel and it looks like my only built-in option is Gboard. I'm guessing they removed it.
There is SwiftKey, on the hand, that does those kind of annoying corrections a couple of times, remembers your choice, and does them no more. It's been a long time since I've seen a 'fir' with SwiftKey.
This is in the european market, though, I don't know if it's configured differently for different markets.
On google the query "samsung turn off keyboard autocorrect" provides links such as https://www.androidcentral.com/how-turn-and-autocorrect-sams... which may or may not be relevant to you.
Someone else suggested Swift keyboard, so I'm going to give that a try.
Google wants you to be mainstream now. If everyone thinks the same and wants the same things -- even if contextually as a member of one of a couple of hundred disparate "marketing cohort" categories -- it will be far easier to target advertising to you. It's in Google's interest for you to conform now. Be easy to categorize. Be easy to predict. So think the same as the members of your peer group, so they can sell hyper-targeted advertising to other corporations. (Have you noticed that social media tends to motivate you to conform?) Google has no use for the long tail anymore -- no use for quirky and inscrutable scenes and subcultures. Instead, it now has the cultural power to transform you (the product) into an even better product.
Remember: 1. If you're not paying for the service, you are not the customer. You are the product. 2. Given sufficient quantity and concentration, Power corrupts. Always.
However, the observation of this article, and my observation as well, is that Google isn't currently capable of parsing very individual quirks. Rather, Google is able to place you into one of a number of highly conformist boxes. They don't have to understand you as an individual. They just have to 'box' you more effectively than their competitors.
There is nothing in the market or otherwise emergent in the nature of data and such categorization which fundamentally motivates Google to be able to parse anyone's quirks or understand the essence of a scene or artistic movement. If Google can gain a competitive advantage by creating a number of honeytrap doppelgangers which draw people away from the long tail and sequester them into un-creative, imitative, and highly conformist boxes, then so much the better for them.
https://www.youtube.com/watch?v=R9OHn5ZF4Uo
http://lesswrong.com/lw/l8/conjuring_an_evolution_to_serve_y...
In much the same way, I find that recommendation engines come up with annoying pale imitations of bands/musicians I like. I also wonder why authoritarianism seems to spread so effectively across social media, and why certain authoritarian movements seem to get such ready support from within Google and various social media companies. It's because, as a product, conformist/authoritarian screechers are more easily herded, replicated, categorized, and packaged than real individuals who think for themselves and apply principles.
In other words, the internet is becoming more of a collective and caters less to the individual.
If you have a special interest, then who cares? You should just adopt more normal hobbies. If you have a unique political viewpoint, then get over it and join one of the major parties. If you are oppressed, then it's fine as long as 95% of the people are content (don't worry, we'll carve out a few protected classes so that the pictures still look diverse).
No. _Google_ is being optimized for casual users, and using _Google_ is no longer a skill you can improve to create a path towards a more meaningful web experience.
BTW. this is my go-to example for the "piracy is a service problem" -- this is 100% Disney IP for the fans of two billion dollar movies and you can't buy it as a poster. https://images.moviepilot.com/image/upload/c_fill,h_470,q_au... So I went on eBay, found a custom poster printer service and got it printed and shipped for 12.48 USD: https://i.redd.it/x4mmmvayunbx.jpg I would've been glad to pay double, triple for an official version but no. You just can't buy it.
Showing results for "the force awakens"
Search instead for "the furce awakens"
It seems that context-sensitive search is a curse as well as a blessing. Wikipedia at least offers you disambiguation pages; perhaps a search engine should let you pick a "domain" to prioritise search in.
Is it this thing at the very top of the search results?
First result: https://littlefreelibrary.org
Search: Box library
I'd bet it's probably trained to maximize clicks on ads...
Just like youtubes algorithm is trained to maximize viewing time (and thus ads you see), instead of showing you videos you'd actually enjoy more.
How would youtube quantize how much enjoyment you're having, if not by tracking viewing time?
As an aside, when watching youtube on my TV, I don't think there's a way to thumbs up or down videos. Even on my PC, when the video is in fullscreen, there's no way to thumbs up or down either.
I bet long-tail queries, while capable of carrying very targeted ads with a high CTR, are just too rare and thus less profitable. Likely much more people have basically no clue and formulate queries approximately, and very few query precisely in improbable ways, so getting the majority to the results they "meant" gives better financial results.
To tell the truth, you still can enter the "verbatim mode" using a menu, and try to find that improbable cluster of words. It's an advanced feature now, requiring a bit of digging, but it's there.
Re: long-tail movements and switching contexts between work and home, I wonder if a better example isn't much better user or persona management. Every person is interested in more than just one thing, and can conceivably be looking for the same word in two contexts. Down below, a multi-lingual speaker has given up on predictive keyboards.
What if we could enable users to switch personas/contexts as intuitively and easily as people code-switch in real conversations? Setting up the profiles would be messy and cumbersome at first which probably kills the idea dead in my hands. I'm not knowledgeable enough about psychology or machine learning to figure out if that could be solved automatically.
It would be fine it was work + home persona. I feel though that it woild end up like work local branch + work parent company report + work programming + home family member + home personal hobbies + home grand parent’s health
For context I already have two IDs for work and private stuff, I still hit a lot of barriers on Google search and ended up in ddg, using the location switch when needed.
There has to be an additional reason though, otherwise they could've just put the AI "enhanced" above the fold, and append the actual accurate results below it, for the people doing research.
Example query: +vbs array ~loop "magic number"
If Google does not find anything it will remove some parts so that you will get any results at all
And even written "search term" doesn't work consistently anymore, Google thinks to know better and semi-randomly omit words and show results no one asked for.
They even dropped tilde :(
But you might be able to achieve the same things with their advanced search; https://support.google.com/websearch/answer/2466433?hl=en
I feel this is more a case of Google thinking it knows better.
I have a similar problem with ebay, if I'm looking for an "HP Z430" then I'll get pages of other items where Z430 isn't even in the description.
I can see some value in searching for alternative spellings and related items but there should still be the ability to be exact in your search terms.
The reliance on advertising revenue models means that all such Web properties morph into being essentially adversarial attention traps against users.
I would pay $50/month in a heartbeat for access to a no-ads, deterministic, guts-openened-with-API Google type engine (even if rate limited at that price to some high-human volume of usage).
Never. And definitely not at USD 50 per month. That is a huge amount for this, although I suspect you're a pretty rare customer and/or exaggerating. Broadband or Mobile service, Satellite TV, etc. all have packages that cost about this much. Even a magazine subscription is a fraction of this amount, Netflix is only about 5-10 dollars a month, isn't it? I could see people paying 10% of the Netflix (ad free TV) charge for an ad free search engine maybe so USD 0.50 to 1.00 per month, 1% of your suggestion...
I think the biggest irony is that the web allows for more adoption of long-tail movements than ever before, and Google has gotten significantly worse at turning these up. I assume this has something to do with the fact that information from the long tail is substantially less searched for than stuff within the normal bounds.
This is a nightmare if you have any hobbies that share a common phrase with a vastly more popular hobby, and is especially common when it comes to tech-related activities. I use Linux at home, and I program VBA at work. At home Linux is crossed out of most of the first few pages, and I just get a ton of results about Windows, and at work VBA is crossed off and I get results about VB6 and .NET.
Completely. Useless.
I can only imagine this has something to do with their increasing reliance on AI, and the fact that the AI is probably incentivized to give a correct response to as many people'above the fold' as is possible. If 95% of people are served by dropping the specifically-chosen search term, then the AI probably thinks it's doing a great job.
It seems like the web is being optimized for casual users, and using the internet is no longer as skill you can improve to create a path towards a more meaningful web experience.