The YouTube search team works closely with the web search team and uses many of the same technologies. However, they optimize for different metrics and goals.
I really love outlook as an email client, at least on Windows. On a mac it still needs some work. It's getting better as more features are added.
Must admit though, searching is always a bit of a pain, sometimes I can't seem to find emails that I know for sure are there, and after some browsing I'm able to find them.
At least searching chat history used to work well back in the XMPP era of Google Talk. It feels like it the search quality strongly declined in last years.
Google seems to favour desktop search over mobile for some reason. I can't search chats from my phone (but I can from desktop), I can't even set up filters from mobile.
There are some search tricks you can employ from the desktop to help finding stuff easier (e.g. narrow by date) so I'd recommend you use these on desktop to try to locate what you need. When you find it, apply a tag to make finding it easier in the future.
What? I love Gmail and Hangouts search. I'll be very upset when they get rid of Hangouts because it's so convenient to search. What's frustrating is that you can't search Hangouts on mobile without going to the desktop Gmail website.
What's so hard about searching chat history? If you just use in:chat, that works fine for me. For anything else, just using in:all usually surfaces threads with the relevant terms, although you can also filter out stuff in the social and promotion tabs if you don't want results from mailing lists or whatever.
There are a few advanced search operators that should be there that are missing, like finding the threads with the most number of messages. And the APIs are missing some functionality also, but overall it seems to work pretty well.
Another example: why is Google Drive search so bad?
It seems like there is rich data to train a model on: graph of shares between the org, comment activity, time open, collective time spent writing, etc.
It just matches terms literally and shows matching emails in chronological order. It doesn't do ranking and non-literal matches like web search does. I think this is the right way of doing it.
Also, the gmail team has to work with a lot of restrictions because of privacy. They are not allowed to look at user's email ever. That would make it hard to do ranking.
Can you give examples of why its bad? While it isn't as contextual as it can be like web-search, I haven't found it to be "bad". Slack's search is what I call bad
I'm not a search engineer, but here's my understanding.
Gmail search isn't bad, it just is using a less complete set of signals to intuit what you're searching for. It's more of a "garbage in, garbage out" model than a "we can intuit what you want based on how millions of other people have reacted to the results of this query." That's because you're the only one searching against this data set (your email).
What's an example of something you've had trouble finding via search? There might be better ways to conduct your search. I've lived for 10+ years with my primary organization strategy being to rely on Gmail search, and have found it extremely effective.
Let me add some obvious failings since people seem surprised: The Android Gmail app clearly does separate searches for emails stored on the device and emails that are only in the cloud. The time for the cloud search is very variable, and if it the connection goes bad then there is no indication to the user. This means the same search will mysteriously return different things depending on the connection quality. And if the user is waiting for the cloud results to appear, there is no feedback whatsoever about whether they have almost arrived or whether gmail has given up.
The fact that is it literal matches only seem inexcusable. Sure, the magically ability to guess what your meant isn't going to be available when you're the only one searching your database and there are privacy restrictions, but even just basic spellcheck or near matches would be extremely useful.
When searching old mails I get the distinct feeling they're gone, never to come back ever again. Yes, data loss. I can't prove it because I don't have those emails stored offline.
30 comments
[ 2.2 ms ] story [ 76.7 ms ] threadMust admit though, searching is always a bit of a pain, sometimes I can't seem to find emails that I know for sure are there, and after some browsing I'm able to find them.
There are some search tricks you can employ from the desktop to help finding stuff easier (e.g. narrow by date) so I'd recommend you use these on desktop to try to locate what you need. When you find it, apply a tag to make finding it easier in the future.
https://business.tutsplus.com/tutorials/how-to-search-your-e...
There are a few advanced search operators that should be there that are missing, like finding the threads with the most number of messages. And the APIs are missing some functionality also, but overall it seems to work pretty well.
It seems like there is rich data to train a model on: graph of shares between the org, comment activity, time open, collective time spent writing, etc.
It just matches terms literally and shows matching emails in chronological order. It doesn't do ranking and non-literal matches like web search does. I think this is the right way of doing it.
Also, the gmail team has to work with a lot of restrictions because of privacy. They are not allowed to look at user's email ever. That would make it hard to do ranking.
Maybe this helps: https://support.google.com/mail/answer/7190?hl=en
Gmail search isn't bad, it just is using a less complete set of signals to intuit what you're searching for. It's more of a "garbage in, garbage out" model than a "we can intuit what you want based on how millions of other people have reacted to the results of this query." That's because you're the only one searching against this data set (your email).
What's an example of something you've had trouble finding via search? There might be better ways to conduct your search. I've lived for 10+ years with my primary organization strategy being to rely on Gmail search, and have found it extremely effective.
The fact that is it literal matches only seem inexcusable. Sure, the magically ability to guess what your meant isn't going to be available when you're the only one searching your database and there are privacy restrictions, but even just basic spellcheck or near matches would be extremely useful.