matlab_word2vec_binary_reader
Use `readW2Vbin` to read a pre-trained word2vec word embedding model in the binary format. It assumes that the file is written in the following format.
- The data before the first `0x20` (space) are ascii characters representing the number of vocabularies of the model , while the data between the first `0x20` and the first `0x10` (newline) represent the dimension of the word vector. (e.g.,`[ 51 48 48 48 48 48 48 32 51 48 48 10] ` means 3 milion words embedded into 300 dimensions. )
- The main body, which consists of sequence of word-vector pairs, begins right after the newline character. One word-vector pair consists of a sequence of bytes that represents a word, space (0x20), and a sequence of binary data that represents the embedded vector corresponding to the word in single precision (32bit) format. The length of the vector data is 4bytes times number of dimensions (e.g., 1200 bytes for 300 dimension).
This function was tested with the "GoogleNews-vectors-negative300.bin" from the word2vec web (https://code.google.com/archive/p/word2vec/). It took about a minute to read the 3.5GB file.
Cita come
Toru Ikegami (2024). matlab_word2vec_binary_reader (https://github.com/mathworks/matlab_word2vec_binary_reader/releases/tag/v1.2), GitHub. Recuperato .
Compatibilità della release di MATLAB
Compatibilità della piattaforma
Windows macOS LinuxTag
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!Scopri Live Editor
Crea script con codice, output e testo formattato in un unico documento eseguibile.
Versione | Pubblicato | Note della release | |
---|---|---|---|
1.2 | See release notes for this release on GitHub: https://github.com/mathworks/matlab_word2vec_binary_reader/releases/tag/v1.2 |
||
1.1 | See release notes for this release on GitHub: https://github.com/mathworks/matlab_word2vec_binary_reader/releases/tag/v1.1 |
||
1.0 |