Discriminative learning and the lexicon: NDL and LDL

Teachers: Harald Baayen, Yu-Ying Chuang, and Maria Heitmeier (University of Tübingen)

NDL and LDL are simple computational algorithms for lexical learning and
lexical processing. Both NDL and LDL assume that learning is discriminative,
driven by prediction error, and that it is this error which calibrates the
association strength between input and output representations. Both words’
forms and their meanings are represented by numeric vectors, and mappings
between forms and meanings are set up. For comprehension, form vectors predict
meaning vectors. For production, meaning vectors map onto form vectors. These
mappings can be learned incrementally, approximating how children learn the
words of their language. Alternatively, optimal mappings representing the
endstate of learning can be estimated. The NDL and LDL algorithms are
incorporated in a computational theory of the mental lexicon, the
‘discriminative lexicon’. The model shows good performance both with respect to
production and comprehension accuracy, and for predicting aspects of lexical
processing, including morphological processing, across a wide range of
experiments. Since mathematically, NDL and LDL implement multivariate multiple
regression, the ‘discriminative lexicon’ provides a cognitively motivated
statistical modeling approach to lexical processing.

In this course, we will show how comprehension and production of
morphologically complex words can be modeled successfully with the
“Discriminative Lexicon” model for a range of languages (Hebrew, Maltese,
English, German, Dutch, Mandarin Chinese, Korean, Kinyarwanda, Estonian, and
Finnish). We will discuss the kinds of form and meaning representations that
can be set up, including form features derived from the speech signal for
auditory comprehension and semantic features grounded in distributional
semantics. Furthermore, we will provide a survey of the measures that can be
derived from the model mappings to predict empirical response variables such as
reaction times in primed and unprimed lexical decision, spoken word duration,
and tongue movements during speaking. Finally, participants will receive some
training in using the JudiLing package for Julia. This package provides
optimized code for implementing and evaluating components of a “discriminative
lexicon” for a given language.

Selected readings

Baayen, R. H., Chuang, Y. Y., Shafaei-Bajestan, E., and Blevins, J. P. (2019).
The discriminative lexicon: A unified computational model for the lexicon and
lexical processing in comprehension and production grounded not in
(de)composition but in linear discriminative learning. Complexity, 2019.

Baayen, R. H., and Smolka, E. (2020). Modeling morphological priming in German
with naive discriminative learning. Frontiers in Communication, section
Language Sciences, 1-40.

Chuang, Y.-Y., and Baayen, R. H. (in press). Discriminative learning and the
lexicon: NDL and LDL. Oxford Research Encyclopedia of Linguistics.

Chuang, Y. Y., Kang, M., Luo, X. F. and Baayen, R. H. (to appear). Vector Space
Morphology with Linear Discriminative Learning. In Crepaldi, D. (Ed.)
Linguistic morphology in the mind and brain.

Chuang, Y-Y., Vollmer, M-l., Shafaei-Bajestan, E., Gahl, S., Hendrix, P., and
Baayen, R. H. (2020). The processing of pseudoword form and meaning in
production and comprehension: A computational modeling approach using Linear
Discriminative Learning. Behavior Research Methods, 1-51.

Heitmeier, M., Chuang, Y-Y., Baayen, R. H. (2021). Modeling morphology with
Linear Discriminative Learning: considerations and design choices. Frontiers in
Psychology, 12.

Preprints available at https://www.sfs.uni-tuebingen.de/~hbaayen/publications.html.

University of Groningen logo
theme by teslathemes adapted by Jelmer Borst
theme by teslathemes

adapted by Jelmer Borst