A program for training a Word2Vec model based on news agency corpus (or any corpus of texts given as a set of docunments).
After chosing your language of preference (Java or Python) please follow the instructions below:
-
Java branch:
-
If you prefer to build the image, you must build the Java project using Apache Maven (install Maven if necessary) with
mvn clean packagethen build the dokcer image withsudo docker-compose build. -
For basic configuration see conf/word2vec-default.properties file.
-
Python branch: please have a look at conf/default-conf.yml.
For starting the traning process run the following in shell
$ sudo docker-compose up -d
- Adding more documetation.
For running the program you should have prepared a text file conntaing the docunments where each line designates a (clean) document.