+noir +film -"pinot noir"
(Edit: scale and diversity of input data but also audiences)
Google regularly fails to include all words I searched (even if it's only three or four), often retrieving completely useless results. I doubt it's due to incompetence; I take it as a signal that they're now struggling to match the volume and characteristics of the data they have to ingest to an adequate user experience.
For a large number of people, Google's ability to answer the underlying question, rather than explicitly identify pages where all search terms appear, means it works better. If you think of Google as a way to get answers, this is good.
If you think of Google as a search engine, and particularly if you have historical experience with (and expectations of) search engines, this is very frustrating. And the workarounds of clicking the "must contain" link (or surrounding all of your search terms with quotation marks) are a seemingly unnecessary inconvenience.
As a personal anecdote, I was an early adopter of smartphones (particularly relative to a non-technical audience). So I was excited when I could speak to my phone, then disappointed when I discovered that I had to structure my queries and instructions very carefully.
A few years ago I was on a road trip with a very non-technical friend. We decided to stop for Chipotle. Had it been up to me, I would probably have pulled out my phone, opened Google Assistant (or perhaps Maps directly), and told my phone (speaking as clearly as possible) "navigate to the closest Chipotle" or something similar.
But I was driving, so she just pulled out her iPhone and half-shouted "I want a burrito!" at it. And that worked just fine.
Point being, I had expectations for how things should work based on interactions with earlier iterations of an interface. She didn't.
Search engines being 'about' a concept isn't a new thing. Bewlew's book from 2008 is called "Finding Out About". The dream is that the search engine can work out what a document is about, and what I'm thinking about based on the content of the document / query, and match them up.
The new thing in your example is that Google has gone beyond documents into burrito restauraunts, but it's not such a huge leap.
Maybe new adavances have brought new algorithms that are somehow better at finding and modelling those abstractions so the search engine is no longer a recognisable vector space model with predictable proxies. Even if that _is_ the case, they should be able to answer a query I have made, on my own terms.
Google really needs to develop a "pro mode" search engine that works for this use case. I get the need for an "answers engine" for less savvy users and more casual use cases, but it's a massive company. It can afford to execute two products in its core competency (rather than umpteen messaging apps that it will kill, along with a lot of other useless and/or doomed stuff).
It keeps the power users on the site so they wont have to look for an alternative. Power users if they find something better might influence no power users to the other site
Plus, they're inadequate.
Google returned exactly ZERO results about TPUG. All of the results were about dogs.
That's either very quick turnaround to fix, or a deeper mystery!
Two people sitting at machines next to each other can perform the same search and get different results. It's what Google's spent billions of dollars on.
I figured it was due to the user base of 2019 being very different than that of 2010, and Google adapting to the fact that most of their users aren't technology literate and cannot formulate clear search queries, so they just try to guess what might be of interest to them.
That's fine, it's their business, but it makes the virtual monopoly even more painful.
The wealth of user traffic is also what no other search engine can replicate, due to Google's market share in web searches.
When Google receives a search query, it first broadens the search phrase (see [0]). The user's clickstream and search refinements are helpful in both training the model for doing the broadening, and then weighting the search contexts, for narrowing down what should actually be displayed on the front pages.
[0] https://www.link-assistant.com/news/keyword-refinements.html
Exactly! Search engine performance can be assessed by measuring precision and recall [0]. Full text search engines have really high precision. Additionally, when the user has been socialized with full text searches, they've built a model of how the search engine works ("it will find documents which contain my search phrase"), so false negatives are perceived to be less severe, as they can be readily explained by the model. "Ah, this document about helicopers contains 'Apache', no wonder it's in the results. I'll add 'webserver' to narrow it down" (And experienced users will already start off with all necessary key terms).
While full text search engines have high precision, they also have bad recall. This can be improved, but there is a tradeoff when tuning the algorithm: to increase recall, the search context is broadened. That necessarily decreases precision as well, because there is no way the search engine is always correct when adding context. Also, when at first all documents on the frontpage at least contained the search term, now there is not even a good explanation why some documents were retrieved. And the more precise the query itself (something we learned by using full text searches) the higher the probability of misclassification, and the worse the effects of broadening. The relevant results are somewhere in the list, but now every second result on the frontpage is from the wrong bucket. And with no explanation, those false positives weight heavy for us users from the old days.
[0] Precision is the probability that a random document in the result set is relevant. Recall is the probability that a random relevant document is in the result set.
No.
That is awfully generous of you ...
Google shows us what it shows us to maximize advertising revenue. They need you to keep clicking and generate hits on adwords-encumbered websites. Showing you zero (or one or a handful of) results for your search query is counter to this goal.
Showing you a batch of results with no adwords-encumbered sites in it is also counter to this goal.
I don't think they're struggling with anything at all - they are optimizing for paid clicks and precise search results is, at best, a very distant second priority...
"noir" "film" -"pinot noir"
I wonder whether they'll revert back now that google+ is dead.
This really confused me, because I was trying to find something specific, and it found a few emails, so I read them, and then was confused that they didn't actually take about the specific thing I was trying to remember. Then I realized that they I included one of the words from my string.
In this case, a close match was completely useless, and ended up wasting my time reading irrelevant results. A message saying "we didn't find that, but here's a few close matches" would have been more helpful and avoided wasting my time.
The reasons in this thread about non precise search results are half the reason I stopped using Google Search about a year ago. I get why they've done it, but I don't like it.
I either use duckduckgo or google in another browser that I don't normally use.
Give it a try - I run CookieAutoDelete so it is 'theoretically' a clean search each time if that makes any difference.
I remember trying to google something about Sufism and its relationship to mysticism and the occult...and google brought up a bunch of results from right wing conspiracy websites claiming that Islam was related to the "New World Order".
[0] https://searchenginewatch.com/sew/news/2126346/google-introd...
It was my first exposure to how much work could be saved by working efficiently with unstructured data! Of course, Google took this to a whole other level, realizing that for common users queries themselves should be treated as unstructured data! Learning as much as you can from how people already express themselves is one way to write the future, it turns out...
afaik putting a term in double quotes does the same thing but I am unsure if the implementation is really the same, the effect seems to be
I can't. The plus operator meant the the word you applied it to must be in the search result. Quoting the word does not do this for me.