581–590 of 639 posts
Earlier quoted context omitted.
As a postdoc in computational linguistics, my go-to example for talks is asking Siri not to show me the weather.
... Siri doesn't show you the weather by default. I know your point is about shallow parsing, but there's a reason it still kinda works.
Earlier quoted context omitted.
Like, when was the last time this was considered ideal? “Yes, I would like an unstriped dress shirt please” “How about this striped shirt?” “No thank you, I would like an unstriped dress shirt please” “I have some lovely jogging pants” “Ok, I need to be clear here, I would like a dress shirt that has no stripes” “Can I interest you in a white undershirt? People who buy dress shirts usually buy undershirts” ....
I think if you were visiting a personal tailor, or perhaps talking to a shop assistant, you might get something akin to that. T: I think pink would look good on you, and it's very fashionable right now. You: Just bring me some yellow shirts to try. T: Oh, I got these, and brought this pink one anyway; try it! But, of course Google isn't making fashion suggestions. But then, ... the tailor might also be just trying to…
They can certainly also bring some stock to shift, or offer suggestions while I’m trying something on, but if they aren’t listening when I make a direct request or when I clearly say no, then they aren’t really there for me, their customer.
=> "shirt -stripes" works pretty well on google at least
Earlier quoted context omitted.
Down voted because this is just a lie. "Kind person" - pictures of men women, children, of all ages and colors. "good person" - Mostly pictures of two hands holding. No clear bias towards women at all. If anything, more of the hands look "male". "Bad person" - Nearly 100% cartoon characters Absolutely ridiculous that you would take the time to write up such fake nonsense.
Google searches are not reproducible, different users can get different results on the same query.
Following the stereotype content model theory I would likely get a pretty decent prediction of what kind of culture and group perspective produced the data. You could also rerun the experiment in different locations to see if it differ.
This problem is known as "attribution" - you have a "no" or "without" in the sentence, but you don't know where it belongs. One could (and one does) argue that the problem cannot be solved with statistical methods (ML), especially not in any domain where accuracy is required, such as medical recored analysis: "no evidence of cancer" and "evidence of no cancer" are very different things. Zooming out, the language fiel…
Earlier quoted context omitted.
Every human understands that phrase and yet the AI doesn't. That's the gap that has to be fixed.
You have to realize that search is not AI. It's pattern matching. And the string "shirt without stripes" matches really good with "shirt with stripes". Levenshtein distance is 3. If vendors would use the term "shirt without stripes" than it would match great, but they call it "plain shirt".
Google advertises using BERT natural language models
https://blog.google/products/search/search-language-understa...
> ... but they call it "plain shirt".
Or polka dotted :)
Why is this a Github repo? I can't get past the abuse of a git repository.
Earlier quoted context omitted.
I've noticed this too, and frequently wonder why there aren't new and better search startups...
There are, like DuckDuckGo. But the first complaint is usually "their results aren't as good as Google" and that's because Google in reality still gives better results because of their (lack of) privacy. People want better results but don't want to be tracked, and those things are in opposition to each other.
Alternative hypothesis: people only have had Google as reference for years, which means that Google represents "reality" to them. Anything that looks even slightly different is therefore worse.