An exploration of the encoding of grammatical gender in word embeddings
This work addresses the problem of understanding linguistic representations in NLP for researchers, but it is incremental as it applies existing methods to analyze gender encoding.
The study investigated how grammatical gender is encoded in word embeddings across Swedish, Danish, and Dutch, finding overlap in encoding and that adding contextual information harms classifier performance while removing morpho-syntactic features like articles dramatically reduces accuracy.
The vector representation of words, known as word embeddings, has opened a new research approach in linguistic studies. These representations can capture different types of information about words. The grammatical gender of nouns is a typical classification of nouns based on their formal and semantic properties. The study of grammatical gender based on word embeddings can give insight into discussions on how grammatical genders are determined. In this study, we compare different sets of word embeddings according to the accuracy of a neural classifier determining the grammatical gender of nouns. It is found that there is an overlap in how grammatical gender is encoded in Swedish, Danish, and Dutch embeddings. Our experimental results on the contextualized embeddings pointed out that adding more contextual information to embeddings is detrimental to the classifier's performance. We also observed that removing morpho-syntactic features such as articles from the training corpora of embeddings decreases the classification performance dramatically, indicating a large portion of the information is encoded in the relationship between nouns and articles.