latent semantic factoring

• Latent Semantic Analysis: Latent Semantic Analysis, LSA

• Latent Semantic Indexing or Latent Semantic Index: Latent Semantic Indexing, LSI.

• Sparse matrix with each row for each word and each column for each document

  • Use tf-idf to weight each component
  • Do a singular value analysis on this matrix.
  • If we choose k largest singular values, we can embedding with the smallest error into k dimensions - Frobenius Norm.
  • Equivalent to learning a function for obtaining a word tf-idf vector from a document ID in an Autoencoder-like network with k hidden layers and 1 layer

---

This page is auto-translated from /nishio/潜在意味解析 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.