Hi Amos,
You can create an embedding matrix for the 'GLoVE' embeddings by initializing a matrix of size 400K × 300 initialized with 'zeros': https://www.mathworks.com/help/matlab/ref/zeros.html Corresponsingly each line can be read and stored (only the numeric part) in the matrix, discarding the word. As the file is in the text format, for storing the word vectors 'str2double':https://www.mathworks.com/help/matlab/ref/str2double.html can be used to convert the text to numbers. Each line in the file looks like this:
the 0.04656 0.21318 -0.0074364 -0.45854 ...
Overall after reading each line the corresponding vector can be stored as follows:
fid = fopen('glove.6B.300d.txt', 'r');
embeddingMatrix = zeros(400000, 300);
for i = 1:400000
line = fgetl(fid);
tokens = strsplit(line);
embeddingMatrix(i, :) = str2double(tokens(2:end));
end
fclose(fid);
You can also refer to following other useful documentation pages by MathWorks:
Hope this helps!
