如何计算预训练词嵌入的相似度

Question

我想知道R中预先训练过的嵌入向量中最相似的词。例如:类似于“啤酒”的词。为此，我在http://nlp.stanford.edu/data/glove.twitter.27B.zip上下载了预先训练过的嵌入向量，并应用了以下代码：

源代码：

glove_dir = "~/Downloads/glove.6B"
lines <- readLines(file.path(glove_dir, "glove.6B.100d.txt"))
embeddings_index <- new.env(hash = TRUE, parent = emptyenv())
for (i in 1:length(lines)) {
    line <- lines[[i]]
    values <- strsplit(line, " ")[[1]]
    word <- values[[1]]
    embeddings_index[[word]] <- as.double(values[-1])
}
cat("Found", length(embeddings_index), "word vectors.\n")

embedding_dim <- 100
embedding_matrix <- array(0, c(max_words, embedding_dim))
for (word in names(word_index)) {
  index <- word_index[[word]]
  if (index < max_words) {
    embedding_vector <- embeddings_index[[word]]
    if (!is.null(embedding_vector))
      embedding_matrix[index+1,] <- embedding_vector
  }
}

但我不知道如何得到最相似的词。我找到了示例，但没有工作，因为嵌入向量的结构是不同的

find_similar_words <- function(word, embedding_matrix, n = 5) {
    similarities <- embedding_matrix[word, , drop = FALSE] %>%
    sim2(embedding_matrix, y = ., method = "cosine")
    similarities[,1] %>% sort(decreasing = TRUE) %>% head(n)
}
find_similar_words("beer", embedding_matrix)

R中预训练词嵌入的相似度计算

Answer 1

一种解决方案可能是使用text-package (www.r-text.org)。

# for installation guidelines see:  http://www.r-text.org/articles/Extended_Installation_Guide.html

library(text)
text_example <- c("beer wine nurse doctor")

text_example_embedding <- textEmbed(text_example, contexts = FALSE)

word_ss <- textSimilarityMatrix(text_example_embedding$singlewords_we)

# The order of the words has changed, so name them according to how they appear in the output. 
colnames(word_ss) <- text_example_embedding$singlewords_we$words
rownames(word_ss) <- text_example_embedding$singlewords_we$words
round(word_ss, 3)