Katrien Beuls

Literature survey on computational learning of construction grammars

Katrien Beuls

The article, published in Constructions and Frames earlier this week, brings together prior work on the computational learning of form-meaning pairings, including – but not limited to – experiments that use Fluid Construction Grammar. The article aims to synthesise the variety of methodologies that have been proposed to date and the results that have been obtained. Moreover, it identifies those parts of the challenge that have been successfully tackled and reveal those that require further research. Finally, it offers a comprehensive roadmap which can help to boost and streamline future research efforts on the computational learning of large-scale, usage-based construction grammars. The article can be read through the publisher’s web page (open access): https://www.jbe-platform.com/content/journals/10.1075/cf.23026.dou

Bringing computational construction grammar into your classes and research

Katrien Beuls

Paul Van Eecke and Katrien Beuls are organising a tutorial at the 11th International Conference on Construction Grammar in Antwerp on 17 August 2021. During the tutorial, we will walk you through the basics of computational construction grammar (CCxG), with a special focus on how the main ideas underlying construction grammar can be implemented using FCG. The tutorial will alternate between theory, hands-on exercises from the textbook and demonstrations of more advanced case studies. It targets in particular lecturers in construction grammar who wish to include CCxG into their courses, as well as scholars who would like to learn CCxG and use it in their research. For more information, please check the tutorial’s webpage.

A computational construction grammar approach to semantic frame extraction

Katrien Beuls

We are happy to announce that our latest FCG paper “A Computational Construction Grammar Approach to Semantic Frame extraction” has now been published in the Linguistics Vanguard.

The paper describes a novel approach to extracting semantic frames from texts, with a case study on extracting frames of causation from newspaper articles. The computational construction grammar approach yields a word-level F1 score of 78.5%, outperforming a commonly used approach based on conditional random fields by 4.5 percentage points.

Visual question answering

Katrien Beuls

Have you always wondered what the role of construction grammar could be in solving AI benchmark tasks, such as for example visual question answering? Our latest paper “Computational Construction Grammar for Visual Question Answering” published in the Linguistics Vanguard shows how we were able to write a construction grammar that could parse natural language questions into directly executable queries that can be used to retrieve information in images. Find out more about how we did it in the paper or check out the web demonstration.

Two open PhD positions in computational construction grammar

Katrien Beuls
There are two open PhD positions in the field of computational construction grammar and artificial intelligence. Apply now!
1) Acquiring domain knowledge through natural language dialogue (VUB AI Lab)
 
The selected candidate will work on a project that investigates how gaps in domain knowledge of either a human or an intelligent system can be identified, and filled through natural language dialogue. For doing so, he or she will need to combine symbolic techniques from computational construction grammar and dialogue modelling, with the goal of building a conversational agent that (i) can interact naturally on both the grammatical and discourse level, (ii) reason about the knowledge that needs to be acquired either by the human or by the agent and (iii) integrate the acquired knowledge into its knowledge base for later reuse.  
2) Learning computational construction grammars (University of Leuven & imec)
As a doctoral researcher, you will investigate how construction grammars can be automatically learned by a computational entity (e.g. an autonomous agent), allowing it to communicate in its native environment. You will use a variety of machine learning techniques, ranging from deep neural networks to inductive logic programming. You will set up multi-agent simulations in which a population of autonomous agents makes use of these techniques to establish an effective and efficient communication system.

BJL launch event

Katrien Beuls

We are proud to announce the 30th issue of the Belgian Journal of Linguistics on Computational Construction Grammar and Constructional Change. Please join us for two keynotes and a festive reception on Monday 19 December 2016, at 4pm, in the Royal Library of Belgium in Brussels.

More information on the program and the registration (free but mandatory) can be found on the event’s webpage.

AAAI Spring Symposium, Stanford 27-29 March 2017

Katrien Beuls

Computational Construction Grammar and Natural Language Understanding

CALL FOR PARTICIPATION

This symposium is primarily targeting researchers in natural language processing/understanding and computational linguistics, but will have broader appeal to the larger AI community, specifically researchers in knowledge representation, man-machine interaction, and machine learning. The symposium focuses on recent TECHNICAL advances in computational formalisms for construction grammar, machine learning of construction grammar, and use of construction grammar in natural language understanding and production. The presentation and demonstration of working systems is highly encouraged.

Areas of interest include:

1. What unifies constructional approaches to language?

2. Formalisms and implementations of construction grammar (CxG)

3. Constructicons, corpora annotated for CxG, and statistical construction learning.

4. Construction grammar learning by physical robots.

5. Deep Natural Language Understanding (NLU) and Production.

6. Relations of CxG to the brain, development, and evolution.

7. Application efforts.

Paper submissions may take the form of long papers (4-8 pages in AAAI format, including references) for oral presentation or short papers (up to 4 pages in AAAI format, including references) for poster presentation. Evidence of a working implementation, e.g. through a web-accessible demonstration, is highly encouraged. Individual submissions for oral or short papers are due by OCTOBER 28 and authors will be notified by November 29. There will be one or more hands-on interactive sessions where participants can show implemented CxG systems.

Suggestions for panels may be submitted by October 1 to either steels@arti.vub.ac.be or feldman@icsi.berkeley.edu

All papers for the symposium will be collected and made into an AAAI technical report, which will be distributed to attendees electronically and included in the AAAI Digital Library.

Symposium conveners: Jerome Feldman and Luc Steels

Additional members of program committee: Katrien Beuls, Adele Goldberg and Nancy Chang.

More information: https://ai.vub.ac.be/aaai-construction-grammar-2017/