Vectors and Linguistics: a workshop on word embeddings

Date 23 March 2018 logo leiden 
Time 09:00 - 17:00 hrs.
Location Lipsius Building
Cleveringaplaats 1
2311 BD Leiden
Room Lipsius 127 / 227

 

Vector representations and word embeddings have become increasingly popular in linguistic research. Representing words in a continuous vector space where semantically similar words are embedded close to each other, word embeddings have shown to be very powerful, enabling simple mathematical operations on words.

Famously, word embeddings allow for researchers to discover analogies between words. Solving the equation “[king] - [man] = x - [woman]” (read this as “What relates to woman, as king relates to man?”) for x, gives us a new vector. The word vector closest to that newly calculated vector x is indeed the vector for queen.

Representing language as vectors creates all kinds of other possibilities, too, such as automatic one-on-one translation of words and automatic extraction of verbal paradigms, but its possibilities are not limited to the lexical or the morphological domain.

In this workshop we shall first dive into word embeddings in a hands-on tutorial on how to work with word embeddings yourself. In the afternoon there will be a session with lectures on what can be done with word embeddings and vector representations in linguistic research.

We kindly ask you to enroll for the workshop, since there is a limited amount of places for the hands-on tutorial in the morning (18). You can enroll here. Please indicate whether you plan to come for the whole workshop, or only for the morning or the afternoon. Lunch will not be provided.

Coffee and tea will be served, courtesy of the Leiden University Centre for Digital Humanities.

Morning, Lipsius 127

10.00-10.30 Coffee and tea
10.30-12.00 Hands-on tutorial – Suzan Verberne (LU)

Afternoon, Lipsius 227

13.00-13.45 Arianna Bisazza (LU)
13.45-14.30 Simon Šuster (Uantwerp)
14.30-15.00 Coffee and tea
15.00-15.45 Kris Heylen (KU Leuven)
15.45-16.30 Martin Kroon (LU)