Digital begriffsgeschichte: Tracing semantic change using word embeddings

Figures & data

Table 1. Example co-occurrence matrix.

	Car	Motorcycle	Highway	Helmet	Cauliflower
Car	–	2	3	1	0
Motorcycle	2	–	3	3	0
Highway	3	3	–	3	0
Helmet	1	3	2	–	0
Cauliflower	0	0	0	0	–

Figure 1. Continuous bag-of-words and skip-gram architectures (Figure taken from: (Mikolov, Sutskever, et al. 2013)).

Figure 2. The angle or cosine distance between words along two dimensions (latent factors) indicates how close they are in meaning.

Figure 3. Two-dimensional projection of vectors of countries and their capital cities. (Figure taken from: (Mikolov, Sutskever, et al. 2013).

Figure 4. Semantic Shifts of Individual Words in Dutch Newspapers (dashed line indicates the frequency and the solid line refers to cosine similarity.

Figure 5. Nearest neighbors to “abortion” in 1950-1954, 1975-1979, and 1990-1994.

Figure 6. Changes in local neighbors (k = 25) of two target words in embeddings trained on historical Dutch Newspapers.

Mikolov, T., I. Sutskever, K. Chen, G. S. Corrado, and J. Dean. 2013. Distributed Representations of Words and Phrases and their Compositionality. In Advances in neural information processing systems 26, ed. C.J.C. Burges, L. Bottou, M. Welling, Z. Ghahramani and K.Q. Weinberger, 3111–9. Curran Associates, Inc.

 Google Scholar