6 ms·

Those are some really spicy opinions. It would seem that many search experts might not agree. David Tippet (formerly opensearch and now at Github) A great pod

Those are some really spicy opinions. It would seem that many search experts might not agree. David Tippet (formerly opensearch and now at Github) A great podcast with David Tippet and Nicolay Gerold entitled: "BM25 is the workhorse of search; vectors are its visionary cousin" https://www.youtube.com/watch?v=ENFW1uHsrLM https://www.youtube.com/watch?v=ENFW1uHsrLM

dumb1224 2y ago
Agreed. In the 2000s it was all about BM25 in the NLP community. I hardly see any paper that did not mention it in my opinion.
- RA_Fisher 2y ago
  For sure, it’s very popular, just not the best anymore (and actually far from it).
- authorfly 2y ago
  And dependency chaining. But yes, lots of BM25. The 2000s and even 2010s was a wonderful and fairly theoretical time for linguistics and NLP. A time when NLP seemed to harbor real anonymized general information to make the right decisions with, without impinging on privacy. Oh to go back.
RA_Fisher 2y ago
I’m sure Search experts would disagree, because it’d be their technology they’d be admitting is inferior to another. BM25 is the workhorse, no doubt— but it’s also not the best anymore. Vectors are a step toward learning models, but only a small mid-range step vs. an explicit model. Search is a useful approach for computing learning models, but there’s a difference between the computational means and the model. For example, MIPS is a very useful search algo for computing learning models (but first the learning model has to be formulated).
- simplecto 2y ago
  It seems that the current mode (eg fashion) is a hybrid approach, with vector results on one side, BM25 on the other, and then a re-reank algo to smooth things out. I'm out of my depth here but genuinely interested and curious to see over the horizon.
  - authorfly 2y ago
    Out of interest how come you use the word "mode" here?
    - simplecto 2y ago
      because the space moves fast, and from my learning this is the current thing. Like fashion -- it changes from season to season
      - authorfly 2y ago
        Oh right, I just wondered if it was a loan word from German. I am hearing it more and more in English.
  - RA_Fisher 2y ago
    Best is to use one statistical model and encode the underlying aspects of the context that relate to goal outcomes.
- softwaredoug 2y ago
  I don't know a lot of search practitioners who don't want to use the "new sexy" thing. Most of us do a fair amount of "resume driven development" so can claim to be "AI Engineers" :)
  - RA_Fisher 2y ago
    I don’t think it’s realistic to think that software engineers can pick up advanced statistical modeling on the job, unless they’re pairing with statisticians. There’s just too much background involved.

dumb1224 2y ago

Agreed. In the 2000s it was all about BM25 in the NLP community. I hardly see any paper that did not mention it in my opinion.

RA_Fisher 2y ago
For sure, it’s very popular, just not the best anymore (and actually far from it).
authorfly 2y ago
And dependency chaining. But yes, lots of BM25. The 2000s and even 2010s was a wonderful and fairly theoretical time for linguistics and NLP. A time when NLP seemed to harbor real anonymized general information to make the right decisions with, without impinging on privacy. Oh to go back.

RA_Fisher 2y ago

I’m sure Search experts would disagree, because it’d be their technology they’d be admitting is inferior to another. BM25 is the workhorse, no doubt— but it’s also not the best anymore. Vectors are a step toward learning models, but only a small mid-range step vs. an explicit model. Search is a useful approach for computing learning models, but there’s a difference between the computational means and the model. For example, MIPS is a very useful search algo for computing learning models (but first the learning model has to be formulated).

simplecto 2y ago
It seems that the current mode (eg fashion) is a hybrid approach, with vector results on one side, BM25 on the other, and then a re-reank algo to smooth things out. I'm out of my depth here but genuinely interested and curious to see over the horizon.
- authorfly 2y ago
  Out of interest how come you use the word "mode" here?
  - simplecto 2y ago
    because the space moves fast, and from my learning this is the current thing. Like fashion -- it changes from season to season
    - authorfly 2y ago
      Oh right, I just wondered if it was a loan word from German. I am hearing it more and more in English.
- RA_Fisher 2y ago
  Best is to use one statistical model and encode the underlying aspects of the context that relate to goal outcomes.
softwaredoug 2y ago
I don't know a lot of search practitioners who don't want to use the "new sexy" thing. Most of us do a fair amount of "resume driven development" so can claim to be "AI Engineers" :)
- RA_Fisher 2y ago
  I don’t think it’s realistic to think that software engineers can pick up advanced statistical modeling on the job, unless they’re pairing with statisticians. There’s just too much background involved.