The Meaning of Lines

All Headword Pairs
2022
digital print on aluminum
48 x 67 inches

All Headword Pairs Separated by Letter, A–Z
2022
digital prints on aluminum
8 x 11 inches

Lines connect two points, creating a relationship in visual or mathematical space. Headwords are the words located at the top of a page that help us alphabetically navigate reference texts like encyclopedias and dictionaries. To create this series, Swanson used word vectors to measure and map the difference in meaning between headword pairs, plotting the difference as a line in space. To work properly, word vectors require vast amounts of text to use as training data, but these vectors were unconventionally trained on the dictionary itself. The large print and video animation contain all the headword pairs; the smaller prints contain headwords separated by their initial letter, A–Z. 

Word vectors (or word embeddings) are an attempt to model the semantic meaning of words mathematically. Tis technology is used in text-to-speech applications, spelling and grammar-checking applications, predictive text, and artificial intelligence tools such as ChatGPT.

Word vectors are based on the premise that words with similar meanings are frequently found near each other in a given text. They are created by training a neural network on a large collection of text, or a corpus. The training model measures the distance between each word in the corpus. These distances are used to calculate the probability of finding a given word near another word. These probabilities are then used to create an n-dimensional model of the meaning of the words from the original corpus.