Nlp on Constantin Orasan's website

Christmas market in Craiova

Sun, 07 Dec 2025 00:00:00 +0000

A few weeks ago, I had the pleasure of being an invited speaker the 3rd edition of the Conference on Recent Advances in Digital Humanities. The event was excellent, as expected, and my talk on LLMs and accessibility was warmly received. The real surprise, though, was discovering the city of Craiova and its Christmas market.

Craiova is quite far from my hometown, so I never managed to visit it while I was living in Romania. After accepting the conference invitation, I learned about the Christmas market and wasn’t sure what to expect. It turned out to be remarkable. It was voted the best Christmas market, and after seeing it, I completely understand why. There is even a flying Santa Claus. The market was crowded every evening, but I still managed to capture a few 360-degree photos of the main square.

EmpASR at CIOL 2025

Wed, 26 Mar 2025 00:00:00 +0000

On 15 March 2025, I had the opportunity to present the EmpASR project at the CIOL Interpreters Day 2025 alongside my colleagues Sabine Braun and Diana Singureanu. The title of the presentation was Pros and Cons of AI and Technology in interpreting: A research-informed perspective from the University of Surrey. Despite the fact that we were able to present only preliminary findings from our interviews, the presentation was very well received and generated quite a bit of discussion. The slides of our presentation are available below.

Tree of thoughts

Thu, 21 Sep 2023 00:00:00 +0000

Today, at our NLP Café, I presented

Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L. Griffiths, Yuan Cao, Karthik Narasimhan (2023) Tree of Thoughts: Deliberate Problem Solving with Large Language Models. arXiv:2305.10601

Some of the participants asked for the slides, so here they are. I should mention that I prepared the presentation is a hurry, so it contains many chunks copied from the original appear.

While preparing my presentation, I came across this YouTube video which discussed the paper. In contrast to some other presentations which referred only to the examples in the paper, this video shows how you could simulate Tree of thoughts with a single prompt. I am not sure who came with the idea first, but it is also discussed in Hulbert (2023).

GenAI the next step ...

Sat, 08 Jul 2023 00:00:00 +0000

GenAI: the next step in the evolution of translation?

This is a keynote speech given at the International Conference on Human-Informed Translation and Interpreting Technology (HiT-IT 2023) on 8th July 2023 in Naples, Italy

Slides

Your browser does not support embedded PDFs. You can download the file here.

GenAI and the Future ...

Wed, 28 Jun 2023 00:00:00 +0000

GenAI and the Future of Jobs: What Do We Need to Know About ChatGPT, BARD & Co?

This is an invited talk at the Symposium on “Organising and Experiencing Work in the Digital Age”, The Future of Work Research Centre, University of Surrey, UK

Slides

Your browser does not support embedded PDFs. You can download the file here.

How to create word frequency lists in Python

Thu, 07 Nov 2019 00:00:00 +0000

A few weeks ago I introduced my students to the notion of dictionaries in python. The obvious way of teaching them, especially in the context of my module, is to use dictionaries to create word frequency lists. In order to make the students understand better why dictionaries are useful, we discussed other ways to produce frequency lists which use only lists. This prompted me to try to think as many ways as possible to produce frequency lists. I also wanted to see how feasible is to use these methods.

Word embeddings for aggression identification

Tue, 25 Sep 2018 00:00:00 +0000

The First Shared task on Aggression Identification was organised in conjunction with the First Workshop on Trolling, Aggression and Cyberbullying. The idea of the shared task was fairly simple. Classify a text in one of the following three categories: Overtly Aggressive (OAG), Covertly Aggressive (CAG) and Non-aggressive (NAG). This means that the task is essentially a standard text categorisation task and an approach based on bag-of-words is a good baseline to start with (neither me, nor the task organisers provided a baseline based on bag-of-words, so I don’t know what is the accuracy of the method).