Who: Dr. Fermín Moscoso del Prado
What: Computational Linguistics in the Industrial Setting
When: June 22, 2021 at 12pm EDT (GMT -4)
Where: Zoom

We are now happy to announce the third talk in our series, “Computational Linguistics in the Industrial Setting,” which will be delivered by Fermín Moscoso del Prado on Tuesday, June 22 at 12pm (EST, GMT -4).

Students coming out of graduate programmes in Cognitive Psychology or Linguistics often have very valuable skills to work in industry in the field of data science. There are now abundant and well-paid jobs in this sector across the world. Among the skills that are crucial for success in this field, it is important to have a good knowledge of statistical analyses and machine learning techniques plus some computer programming skills, typically with Python or R. This talk will include discussion of what additional skills and areas of knowledge can help students wishing to go into the data science industry, including a brief overview of types of jobs, career structure, and nomenclature.

Fermín Moscoso del Prado is the Lead Scientist at Lingvist Technologies OÜ, in charge of developing algorithms for improving human vocabulary learning, automatic construction of language course materials, and general data analysis. He holds an M.Eng. in Computer Science specializing in Artificial Intelligence, a Ph.D. in Linguistics (with focus on computational psycholinguistics), and had post-doctoral training in Cognitive Neuroscience.

The meeting will be held using Zoom. Please note that the meeting is password protected and will require a simple pre-registration. If you know somebody who may wish to join, they may contact us or use the Words in the World website in order to be added to the mailing list at https://wordsintheworld.ca/home/open-office-hours/sign-up-for-open-office-hours/.

Who: Marco Marelli (University of Milano-Bicocca, Milano, Italy) www.marcomarelli.net

What: Compositional effects in the processing of compound words: A computational perspective grounded in linguistic and visual experience

When: Monday January 25, 2021; 2:30-4:20 pm EST

Where: Zoom

Registration: https://mcmaster.zoom.us/meeting/register/tJAlduyqrzMqG9HnqMOA1s0NZLhGj-bf2xtU 

McMaster’s Department of Linguistics and Languages invites you to the next talk in the Cognitive Science of Language lecture series. The lecture will be delivered online by Dr. Marco Marelli. Dr. Marelli is an associate professor of General Psychology at the University of Milano-Bicocca, Milano, Italy. His work focuses on the psychology of language, and in particular on the impact of semantics on word processing and the interface between language and the conceptual system. His more recent research projects combine methods from experimental psychology and computational modelling and are dedicated to compositionality (at the level of both phrases and morphologically complex words) and the interplay between linguistic, emotional and perceptual experience in conceptual processes. He is an associate editor of Behavior Research Methods and a consulting editor of Morphology. 

The talk is free but participants must register. Registration link can be found here:  https://mcmaster.zoom.us/meeting/register/tJAlduyqrzMqG9HnqMOA1s0NZLhGj-bf2xtU

Please make sure to register in advance. For logistic reasons the registration for this event will only be reviewed until 2pm on the event date.  

Abstract:  

Since the seminal LSA proposal (Landauer & Dumais, 1997) distributional semantics has provided efficient data-driven models of the human semantic system, representing word meaning through vectors recording lexical co-occurrences in large text corpora. However, these approaches generate static descriptions of the semantic system, falling short of capturing the highly dynamical interactions occurring at the meaning level during language processing. 

In the present work, I discuss the CAOSS model (Compounding as Abstract Operations in Semantic Space), a first step in this direction that moves from distributional semantics to capture the meaning of compound words (Marelli et al., 2017). 

In CAOSS, word meanings are represented as vectors encoding lexical co-occurrences in a reference corpus (e.g., the meaning of “snow” will be based on how often “snow” appears with the other words), according to the tenets of distributional semantics. A compositional procedure is induced as a weighted sum: given two vectors (constituent words) u and v, their composed representation (the compound) can be computed as c=M*u+H*v, where M and H are weight matrices estimated from corpus examples. The matrices are trained using least squares regression, having the vectors of the constituents as independent words (“car” and “wash”,  “rail” and ”way”) as inputs and the vectors of example compounds (“carwash”, “railway”) as outputs, so that the similarity between M*u+H*v and c is maximized. In other words, the matrices are defined in order to recreate the compound examples as accurately as possible. Once the two weight matrices are estimated, they can be applied to any word pair in order to obtain meaning representations for untrained word combinations (e.g., “snow building”). 

In a series of behavioral experiments, model predictions were tested against psycholinguistic data. CAOSS is shown to mirror evidence related to the processing of novel compounds (Marelli et al., 2017; Günther & Marelli, 2020), and in particular the impact of relational information (Gagné, 2001; Gagné & Shoben, 2007) as well as the “morpheme interference effect” (Crepaldi et al., 2010). Moreover, CAOSS also provides a central contribution to the understanding of semantic transparency in familiar compounds: CAOSS estimates are shown to best characterize the transparency impact in word processing (Günther & Marelli, 2019). Finally, I discuss how CAOSS is not to be considered a “disembodied model”, since one can easily ground it in perception by feeding it images together with text data (Günther et al., 2020). 

The model simulations indicate that compositionality-related phenomena are reflected in language statistics. Human speakers are able to learn these aspects from language experience and automatically apply them to the processing of any word combination. The present model is flexible enough to emulate this procedure, predicting sensible relational similarities for novel compounds and correctly capturing the contribution to semantic transparency provided by compositional operations. The model is also shown to generalize to other kind of data, being able to capture the contribution of perceptual experience in the internal dynamics of compound-word processing. Such evidence directly links linguistic composition to conceptual combination, speaking for the possible role of general-level learning procedures at the foundations of both phenomena. 

CCPTalks is a new series presented by the Centre for Comparative Psycholinguistics at the University of Alberta. Join us for their first presentation, featuring Dr. Jiseung Kim (Alberta), on January 22, 2021!

Title: Individual differences in the production and perception of prosodic boundaries in American English

Date: Friday, January 22, 2021

Time: 9:00am MST (GMT-7)

Location: Zoom (contact ccpling@ualberta.ca for link)

Abstract:
I present the findings of my dissertation which investigated the hypothesis that individuals vary in their production and perception of prosodic boundaries, and that the properties they use to signal prosodic contrasts are closely related to the properties used to perceive those contrasts. A group of native speakers of American English participated in an acoustic study and subsequently an eye-tracking study that examined production and perception of three acoustic properties related to Intonational Phrase (IP) boundary: pause, pitch reset, and phrase-final lengthening. The results showed individual differences to a substantial degree, and offered limited evidence of a production-perception relation: a trend was observed in which individuals with longer pause durations.