Word embeddings

Introduction to Natural Language Processing in R

Kasey Jones

Research Data Scientist

The flaw in word counts

Two statements:

Without stop words:

Additional data:

With a word2vec representation, similar words are clustered closer together within a multidimensional vector space.

¹ https://www.adityathakker.com/introduction-to-word2vec-how-it-works/

library(h2o)
h2o.init()

h2o_object = as.h2o(animal_farm)

Tokenize using h2o:

words <- h2o.tokenize(h2o_object$text_column, "\\\\W+")
words <- h2o.tolower(words)
words = words[is.na(words) || (!words %in% stop_words$word),]

word2vec_model <-
    h2o.word2vec(words, min_word_freq = 5, epochs = 5)

h2o.findSynonyms(w2v.model, "animal")

    synonym     score
1     drink 0.8209088
2       age 0.7952490
3   alcohol 0.7867004
4       act 0.7710537
5      hero 0.7658424

h2o.findSynonyms(w2v.model, "jones")

      synonym     score
1      battle 0.7996588
2  discovered 0.7944554
3     cowshed 0.7823287
4     enemies 0.7766532
5       yards 0.7679787

Introduction to Natural Language Processing in R