Ngram2vec toolkit
Ngram2vec is a toolkit aiming at training word vectors of various properties. Different from other popular word embedding toolkits which are designed for a specific representation method, ngram2vec supports arbitrary context features and models, and allows users to improve existing representation methods with no effort.The powerfulness of ngram2vec comes from its decoupled architecture, which is illustrated in figure below. To be more concrete, the decoupled architecture brings the following advantages:
-
extensibility.
-
reusability.
Arbitrary features
Ngram2vec supports numerous context features. The figure below illustrates some examples. What is more, adding new context features can be an easy task thanks to the decoupled design of ngram2vec. Contribution of more context features is welcomed.