sentence-transformers VS umap

txtai is an open-source first system. Given it's own open-source roots, like-minded projects such as sentence-transformers are prioritized during development. But that doesn't mean txtai can't work with Embeddings API services.

Ok great. My use case is not very specific, but rather general. I am looking for a model that can perform asymmetric semantic search for the languages I mentioned earlier (Urdu, Persian, Arabic etc.). I have also looked into the sentence-transformer training documentation. Do you think it would be a good idea to use the XNLI dataset for fine-tuning? Or maybe you can suggest much better dataset. Furthermore, I am not sure if fine-tuning is suitable for my task. Because my use case is general so I can use already trained model.

The task you are looking for is semantic textual similarity. There are a few models and datasets out there that can do this. I'd probably start with the SemEval2017 Task 1 task description and competition entries here and then work outward from there (using something like SemanticScholar or Papers With Code to find newer state of the art works that cite these models if needed). For what it's worth you might find that Sentence Bert (SBERT) gives good vectors for cosine similarity comparison out of the box for this task.

Check out the sentence-transformers implementation. If I don't miss anything they don't exclude CLS when the pooling strategy is set to 'mean'

Break up transcript into shorter segments and convert segments to a 768 vector array. Use a process known as embedding using our second ML model, UKP Labs BERT’s sentence transformer model.

Not sure what kind of texts you have, but these models have a max sequence limit of 512 (approx 350 words or so). If you're texts are longer than that, consider splitting them up into chunks or creating a summary and taking an embedding of that. Some clustering algorithm may be the way to go here. Here's a bunch of examples. I use agglomerative for my use case.

https://github.com/UKPLab/sentence-transformers/tree/master/examples/training/data_augmentation (Augmented Encoding)

For vector representation you can take the mean and then pca to get the size that you want, but if you have time then use sentence transformers to train a vector representation instead.

UMAP GitHub repository: https://github.com/lmcinnes/umap

Uniform Manifold Approximation and Projection (UMAP) is a well-known dimensionality reduction method along with t-SNE.

A number of excellent topic modeling libraries exist in Python today. BERTopic and Top2Vec are two of the most popular. Both use sentence-transformers to encode data into vectors, UMAP for dimensionality reduction and HDBSCAN to cluster nodes.

As with anything, it depends on the problem. But T-SNE and UMAP are often good.

UMAP distances.py: umap/distances.py at master · lmcinnes/umap (github.com)

I used for this project Flask and OpenAI CLIP. For the vector search I used approximate nearest neighbors provided by spotify/annoy. I used Flask-SQLAlchemy with GeoAlchemy2 to query GPS coordinates. The embedding was done using UMAP.

side note: instead of t-SNE consider UMAP - provides better results (and it's much faster) https://github.com/lmcinnes/umap

Sounds like a job for UMAP https://github.com/lmcinnes/umap ?

I am a fairly experienced python developer/researcher (about 10 years), and have found a bug that breaks all of my intuitions. I am messing with the [UMAP](https://github.com/lmcinnes/umap) repository and trying to add the option to disable some additional features. I've stripped everything from it but have a [quick test that will run my UMAP version and compare the outputs with what the original gave](https://github.com/Andrew-Draganov/probabilistic_dim_reduction/blob/master/umap/nndescent_umap_test.py). Managing my random seeds, same inputs, all that.

I'm currently reading through a project on github, https://github.com/lmcinnes/umap, and in `umap/umap_.py` at line 2287, they have this: