Visualizing Concepts over Time

Translantis Project Digital Humanities Approaches to Reference

Cultures: The Emergence of the United States in Dutch Public Discourse 1890-1990

Carlos Martinez-Ortiz, Tom Kenter, Melvin Wevers, Pim Huijnen, Jaap Verheul, and Joris van Eijnatten

DESIGN AND IMPLEMENTATION OF SHICO

SHICO: SHIFTING CONCEPTS OVER TIME

• The emergence and transformation of concepts, ideas, and thoughts

• How can we study concepts in large digitized corpora?

OUR APPROACH• Data: >600.000 digitized

newspaper issues from the Dutch National Library

• Multi-dimensional word-vector space using Google’s word2vec (word embeddings)

• Semantic and syntactic information representation by geometry (Baroni & Kruszweksi, 2014; Wijaya & Yeniterzi, 2011)

1950 1960 1970

1 model = 10 years40 models for period between 1950-1990

TRACING CONCEPTS• Concept is a network of words

• Find closely related words to seed word based on distance

• One or more words as entry-points into concept

• Weighting based on frequency + sum distance

• Tracing concepts: which words remain and disappear from network

• See: Kenter, Wevers, Huijnen, de Rijke (2015) “Ad Hoc Monitoring of Vocabulary Shifts over Time”

expand tosemantic graphwithsemantic spacefor time t+1

vocabulary at time t

prune

t = t + 1

VISUALIZATION

• ADD SCREENSHOT INTERFACE

THREE TYPES OF VISUALIZATIONS

• Stream Graphs

• Network Graphs

• Vocabularies (lists of words in network)

Non-adaptive Model

Adaptive Model

Propaganda: Adaptive & Non-Adaptive

Propaganda: Adaptive & Non-Adaptive

DOPING: STREAM GRAPH

DOPING: VOCABULARIES

DOPING: NETWORK GRAPHS

INTEGRATION

DO-IT-YOURSELF

• Code / how-to-manual is on Github:

• https://github.com/NLeSC/ShiCo

• Models are also on Github

• Easy to integrate into other tools

Thank you!