# Examining Structure of Word Embeddings with PCA

## TL;DR

This paper investigates the structure of Czech word embeddings across different models using PCA, revealing how linguistic information like POS tags is represented and emphasizing the importance of PCA histogram analysis.

## Contribution

It introduces a PCA-based method to analyze the structure of word embeddings and compares how different models encode linguistic features.

## Key findings

- POS information is present in word2vec embeddings.
- NMT decoder embeddings organize POS information more prominently.
- Histogram analysis of PCA components reveals embedding structure details.

## Abstract

In this paper we compare structure of Czech word embeddings for English-Czech neural machine translation (NMT), word2vec and sentiment analysis. We show that although it is possible to successfully predict part of speech (POS) tags from word embeddings of word2vec and various translation models, not all of the embedding spaces show the same structure. The information about POS is present in word2vec embeddings, but the high degree of organization by POS in the NMT decoder suggests that this information is more important for machine translation and therefore the NMT model represents it in more direct way. Our method is based on correlation of principal component analysis (PCA) dimensions with categorical linguistic data. We also show that further examining histograms of classes along the principal component is important to understand the structure of representation of information in embeddings.