5 ms·

The thing that puzzles me about embeddings is that they're so untargeted, they represent everything about the input string. Is there a method for dimensionalit

The thing that puzzles me about embeddings is that they're so untargeted, they represent everything about the input string. Is there a method for dimensionality reduction of embeddings for different applications? Let's say I'm building a system to find similar tech support conversations and I am only interested in the content of the discussion, not the tone of it. How could I derive an embedding that represents only content and not tone?

adamgordonbell 2y ago
Agreed.. biggest problem with off the shelf embeddings I hit. Need a way to decompose embeddings.
johndough 2y ago
You can do math with word embeddings. A famous example (which I now see has also been mentioned in the article) is to compute the "woman vector" by subtracting "man" from "woman". You can then add the "woman vector" to e.g. the "king" vector to obtain a vector which is somewhat close to "queen". To adapt this to your problem of ignoring writing style in queries, you could collect a few text samples with different writing styles but same content to compute a "style direction". Then when you do a query for some specific content, subtract the projection of your query embedding onto the style direction to eliminate the style: query_without_style = query - dot(query, style_direction) * style_direction I suspect this also works with text embeddings, but you might have to train the embedding network in some special way to maximize the effectiveness of embedding arithmetic. Vector normalization might also be important, or maybe not. Probably depends on the training. Another approach would be to compute a "content direction" instead of a "style direction" and eliminate every aspect of a query that is not content. Depending on what kind of texts you are working with, data collection for one or the other direction might be easier or have more/fewer biases. And if you feel especially lazy when collecting data to compute embedding directions, you can generate texts with different styles using e.g. ChatGPT. This will probably not work as well as carefully handpicked texts, but you can make up for it with volume to some degree.
- joerick 2y ago
  Interesting, but your hypothesis assumes that 'tone' is one-dimensional, that there is a single axis you can remove. I think tone is very multidimensional, I'd expect to be removing multiple 'directions' from the embedding.
  - johndough 2y ago
    You could of course compute multiple "tone" directions for every "tone" you can identify and subtract all of them. It might work better, but it will definitely be more work.
  - jerf 2y ago
    I would say rather that the "standard example" is simplified, but it does capture an essential truth about the vectors. The surprise is not that the real world is complicated and nothing is simply expressible as a vector and that treating it as such doesn't 100% work in every way in every circumstance all of the time. That's obvious. Everyone who might work with embeddings gets it, and if they don't, they soon will. The surprise is that it does work as well as it does and does seem to be capturing more than a naive skepticism would expect.
  - mattnewton 2y ago
    No, I don’t think the author is saying one dimensional - the vectors are represented by magnitudes in almost all of the embedding dimensions. They are still a “direction” in the way that [0.5, 0.5] in x,y space is a 45 degree angle, and in that direction it has a magnitude of around 0.7 So of course you could probably define some other vector space where many of the different labeled vectors are translated to magnitudes in the original embedding space, letting you do things like have a “tone” slider.
    - TeMPOraL 2y ago
      I think GP is saying that GGP assumes "tone" is one direction, in the sense there exists a vector V representing "tone direction", and you can scale "tone" independently by multiplying that vector with a scalar - hence, 1 dimension. I'd say this assumption is both right and wrong. Wrong, because it's unlikely there's a direction in embedding space corresponding to a platonic ideal of "tone". Right, because I suspect that, for sufficiently large embedding space (on the order of what goes into current LLMs), any continuous concept we can articulate will have a corresponding direction in the embedding space, that's roughly as sharp as our ability to precisely define the concept.
loa_in_ 2y ago
They don't represent everything. In theory they do but in reality the choice of dimensions is a function of the model itself. It's unique to each model.
- joerick 2y ago
  Yeah, 'everything' as in 'everything that the model cares about' :)
macNchz 2y ago
Depends on the nature of the content you’re working with, but I’ve had some good results using an LLM during indexing to generate a search document by rephrasing the original text in a standardized way. Then you can search against the embeddings of that document, and perhaps boost based on keyword similarity to the original text.
- joerick 2y ago
  Nice workaround. I just wish there was a less 'lossy' way to go about it!
  - jacobr1 2y ago
    Could you explicitly train a set of embeddings that performed that step in the process? For example which computing the loss, you compare the difference against the normalized text rather than the original. Or alternatively do this as a fine-tuning. Then you would have embedding that optimized for the characteristics you care about.
  - hobs 2y ago
    Normal full text search stuff helps reduce the search space - eg lemming, stemming, query simplification stuff were all way before LLMs.
- mrshu 2y ago
  This is also often referred to as Hypothetical Document Embeddings (https://arxiv.org/abs/2212.10496 https://arxiv.org/abs/2212.10496).
  - adamgordonbell 2y ago
    Do you have examples of this? Please say more!
mrshu 2y ago
Though not exactly what you are after, Contextual Document Embeddings (https://huggingface.co/jxm/cde-small-v1 https://huggingface.co/jxm/cde-small-v1), which generate embeddings based on "surrounding context" might be of some interest. With 281M params it's also relatively small (at least for an embedding model) so one can play with it relatively easily.
nostrebored 2y ago
There are a few things you can do. If these access patterns are well known ahead of time, you can train subdomain behavior into the embedding models by using prefixing. E.g. content: fixing a broken printer, tone: frustration about broken printer, and "fixing a broken printer" can all be served by a single model. We have customers doing this in production in other contexts. If you have fundamentally different access patterns (e.g. doc -> doc retrieval instead of query -> doc retrieval) then it's often time to just maintain another embedding index with a different model.
_pastel 2y ago
You could fine-tune the embedding model to reduce cosine distance on a more specific function.
magicalhippo 2y ago
I've just begun to dabble with embeddings and LLMs, but recently I've been thing about tryin to use principle component analysis[1] to either project to desirable subspaces, or project out undesirable subspaces. In your case it would be to take a bunch of texts which roughly mean the same thing but with variance in tone, compute PCA of the normalized embeddings, take the top axsis (or top few) and project it out (ie subtract the projection) of the embeddings for the documents you care about before doing the cosine similarity. Something along those lines. Could be it's a terrible idea, haven't had time to do much with it yet due to work. [1]: https://en.wikipedia.org/wiki/Principal_component_analysis https://en.wikipedia.org/wiki/Principal_component_analysis

adamgordonbell 2y ago

Agreed.. biggest problem with off the shelf embeddings I hit. Need a way to decompose embeddings.

johndough 2y ago

You can do math with word embeddings. A famous example (which I now see has also been mentioned in the article) is to compute the "woman vector" by subtracting "man" from "woman". You can then add the "woman vector" to e.g. the "king" vector to obtain a vector which is somewhat close to "queen". To adapt this to your problem of ignoring writing style in queries, you could collect a few text samples with different writing styles but same content to compute a "style direction". Then when you do a query for some specific content, subtract the projection of your query embedding onto the style direction to eliminate the style: query_without_style = query - dot(query, style_direction) * style_direction I suspect this also works with text embeddings, but you might have to train the embedding network in some special way to maximize the effectiveness of embedding arithmetic. Vector normalization might also be important, or maybe not. Probably depends on the training. Another approach would be to compute a "content direction" instead of a "style direction" and eliminate every aspect of a query that is not content. Depending on what kind of texts you are working with, data collection for one or the other direction might be easier or have more/fewer biases. And if you feel especially lazy when collecting data to compute embedding directions, you can generate texts with different styles using e.g. ChatGPT. This will probably not work as well as carefully handpicked texts, but you can make up for it with volume to some degree.

joerick 2y ago
Interesting, but your hypothesis assumes that 'tone' is one-dimensional, that there is a single axis you can remove. I think tone is very multidimensional, I'd expect to be removing multiple 'directions' from the embedding.
- johndough 2y ago
  You could of course compute multiple "tone" directions for every "tone" you can identify and subtract all of them. It might work better, but it will definitely be more work.
- jerf 2y ago
  I would say rather that the "standard example" is simplified, but it does capture an essential truth about the vectors. The surprise is not that the real world is complicated and nothing is simply expressible as a vector and that treating it as such doesn't 100% work in every way in every circumstance all of the time. That's obvious. Everyone who might work with embeddings gets it, and if they don't, they soon will. The surprise is that it does work as well as it does and does seem to be capturing more than a naive skepticism would expect.
- mattnewton 2y ago
  No, I don’t think the author is saying one dimensional - the vectors are represented by magnitudes in almost all of the embedding dimensions. They are still a “direction” in the way that [0.5, 0.5] in x,y space is a 45 degree angle, and in that direction it has a magnitude of around 0.7 So of course you could probably define some other vector space where many of the different labeled vectors are translated to magnitudes in the original embedding space, letting you do things like have a “tone” slider.
  - TeMPOraL 2y ago
    I think GP is saying that GGP assumes "tone" is one direction, in the sense there exists a vector V representing "tone direction", and you can scale "tone" independently by multiplying that vector with a scalar - hence, 1 dimension. I'd say this assumption is both right and wrong. Wrong, because it's unlikely there's a direction in embedding space corresponding to a platonic ideal of "tone". Right, because I suspect that, for sufficiently large embedding space (on the order of what goes into current LLMs), any continuous concept we can articulate will have a corresponding direction in the embedding space, that's roughly as sharp as our ability to precisely define the concept.

loa_in_ 2y ago

They don't represent everything. In theory they do but in reality the choice of dimensions is a function of the model itself. It's unique to each model.

joerick 2y ago
Yeah, 'everything' as in 'everything that the model cares about' :)

macNchz 2y ago

Depends on the nature of the content you’re working with, but I’ve had some good results using an LLM during indexing to generate a search document by rephrasing the original text in a standardized way. Then you can search against the embeddings of that document, and perhaps boost based on keyword similarity to the original text.

joerick 2y ago
Nice workaround. I just wish there was a less 'lossy' way to go about it!
- jacobr1 2y ago
  Could you explicitly train a set of embeddings that performed that step in the process? For example which computing the loss, you compare the difference against the normalized text rather than the original. Or alternatively do this as a fine-tuning. Then you would have embedding that optimized for the characteristics you care about.
- hobs 2y ago
  Normal full text search stuff helps reduce the search space - eg lemming, stemming, query simplification stuff were all way before LLMs.
mrshu 2y ago
This is also often referred to as Hypothetical Document Embeddings (https://arxiv.org/abs/2212.10496 https://arxiv.org/abs/2212.10496).
- adamgordonbell 2y ago
  Do you have examples of this? Please say more!

mrshu 2y ago

Though not exactly what you are after, Contextual Document Embeddings (https://huggingface.co/jxm/cde-small-v1 https://huggingface.co/jxm/cde-small-v1), which generate embeddings based on "surrounding context" might be of some interest. With 281M params it's also relatively small (at least for an embedding model) so one can play with it relatively easily.

nostrebored 2y ago

There are a few things you can do. If these access patterns are well known ahead of time, you can train subdomain behavior into the embedding models by using prefixing. E.g. content: fixing a broken printer, tone: frustration about broken printer, and "fixing a broken printer" can all be served by a single model. We have customers doing this in production in other contexts. If you have fundamentally different access patterns (e.g. doc -> doc retrieval instead of query -> doc retrieval) then it's often time to just maintain another embedding index with a different model.

_pastel 2y ago

You could fine-tune the embedding model to reduce cosine distance on a more specific function.

magicalhippo 2y ago

I've just begun to dabble with embeddings and LLMs, but recently I've been thing about tryin to use principle component analysis[1] to either project to desirable subspaces, or project out undesirable subspaces. In your case it would be to take a bunch of texts which roughly mean the same thing but with variance in tone, compute PCA of the normalized embeddings, take the top axsis (or top few) and project it out (ie subtract the projection) of the embeddings for the documents you care about before doing the cosine similarity. Something along those lines. Could be it's a terrible idea, haven't had time to do much with it yet due to work. [1]: https://en.wikipedia.org/wiki/Principal_component_analysis https://en.wikipedia.org/wiki/Principal_component_analysis