Natural Language Processing

Jose Camacho Collados

about me

Jose Camacho Collados

Jose Camacho Collados
Background. I am currently a Lecturer at the School of Computer Science and Informatics at Cardiff University. I am also a UKRI Future Leaders Fellow since February 2021, and run the Cardiff NLP group. Before that I was a postdoc for a year on the FLEXILOG ERC project with Steven Schockaert. Previously I was a Google Doctoral Fellow and PhD student at the Linguistic Computing Laboratory (LCL) of Sapienza University of Rome. My background education includes an Erasmus Mundus Master in Natural Language Processing and Human Language Technology and a 5-year BSc degree in Mathematics. I also worked for a year as a research engineer at ATILF-CNRS in Nancy (France).

Research. I work on various topics in Natural Language Processing (NLP), mainly on the lexical and distributional semantics areas. In this area I've recently written a book (titled Embeddings in NLP) with Taher Pilehvar that can give an overview of the recent trends in distributional semantics and NLP. These last years I have been particularly interested on how relational knowledge is captured in current NLP models (embeddings/language models), and how this plays a role in applications. During my PhD I've also worked on integrating explicit knowledge (mainly from lexical resources) into downstream NLP applications, with a special focus on multilinguality and ambiguity. To this end, I have been collaborating on the BabelNet project and developing knowledge-based sense vector representations (e.g. NASARI and SW2V) to be used as a bridge between lexical resources and text-based applications. We have organized a tutorial at ACL 2016 and a workshop at EACL 2017 on this topic, and a tutorial at NAACL 2018 on the interplay between lexical resources and NLP.

Open data. I strongly believe that well-curated datasets and resources, as well as shared tasks, are key for advancing science. In 2019 we organized the WiC challenge on evaluating context-sensitive representations. This competition was part of a shared task in the IJCAI workshop SemDeep, and is featured in the SuperGLUE language understanding benchmark. We recently extended this effort with the WiC-TSV benchmark. In the past I also helped organize several SemEval tasks in Word Similarity Hypernym Discovery and Emoji Prediction. Check them out as all datasets are openly available and in various languages!
Finally, I have also been working on social media applications recently, for which we also have open datasets (TweetEval), multilingual language models (XLM-T) and cross-lingual word embeddings!

Other. NLP aside, I love travelling and sports. I was raised in Granada, a wonderful city in the south of Spain where I spent the first 20 years of my life. Then, I have been living in large European cities like Paris, Barcelona and Rome, and spent long amounts of time in Seoul. I have also lived in other smaller (but equally charming) cities: Nancy and Besançon (France) and Wolverhampton (UK). I like practising all kinds of sports: football, swimming, tennis, padel, ping pong... and chess (yes, it is also a sport!). I hold the International Master chess title and am currently the top-rated chess player of South Korea.

Teaching and Supervision: You can find more about my current teaching and PhD supervision commitments here.
Note: If you are interested in doing a PhD with me, please read this note to prospective PhD students.


Daniel Loureiro, Alípio Mário Jorge and Jose Camacho-Collados.
LMMS Reloaded: Transformer-based Sense Embeddings for Disambiguation and Beyond. [paper] [data&code]
arXiv preprint arXiv:2105.12449 (2021).
Francesco Barbieri, Luis Espinosa Anke and Jose Camacho-Collados.
XLM-T: A Multilingual Language Model Toolkit for Twitter. [paper] [data&code]
arXiv preprint arXiv:2104.12250 (2021).
Asahi Ushio, Luis Espinosa-Anke, Steven Schockaert and Jose Camacho-Collados.
BERT is to NLP what AlexNet is to CV: Can Pre-Trained Language Models Identify Analogies? [paper] [data&code]
ACL 2021
Daniel Loureiro*, Kiamehr Rezaee*, Mohammad Taher Pilehvar and Jose Camacho-Collados.
Language Models and Word Sense Disambiguation: An Overview and Analysis. [paper] [data&code]
Computational Linguistics (2021)
Francesco Barbieri, Jose Camacho-Collados, Leonardo Neves and Luis Espinosa-Anke.
TweetEval: Unified Benchmark and Comparative Evaluation for Tweet Classification. [paper] [data]
Findings of EMNLP 2020.
Zied Bouraoui, Jose Camacho-Collados and Steven Schockaert.
Inducing Relational Knowledge from BERT. [paper]
AAAI 2020, New York, USA.


August 2021. Taher Pilehvar and I taught a week-long course on "Embeddings in NLP" at the ESSLLI summer school.

February 2021. Just got started with my UKRI Future Leader Fellowship!

December 2020. Our "Embeddings in NLP" book has been published! All information, including a short video tutorial, here.

December 2020. Join us at Cardiff University! Call for five open-ended Lecturer Positions at the School of Computer Science.

November 2020. Hiring! Looking for a 3-year postdoc starting early/mid 2021. All the details on how to apply here.

October 2020. Excited and honoured to be awarded a UKRI Future Leaders Fellowship!

September 2020. Recently received a research grant from Snap Research (w/ Luis Espinosa-Anke and Daniel Loureiro) and started a collaboration for investigating meaning shift in social media.

July 2020. We have founded the Cardiff NLP group, which now has a Twitter account and a website!



Jose Camacho Collados

code / resources

Jose Camacho Collados

- WiC-TSV, a new benchmark extending word to retrieve senses in context.

- Meemi, an open-source implementation to learn cross-lingual embeddings, including pre-trained models.

- WiC, the Word-in-Context dataset for evaluating context-sensitive representations.

- NASARI vector representations for BabelNet synsets and Wikipedia pages (English, French, German, Italian, Spanish).

- SW2V (Senses and Words to Vectors): Word and Sense embeddings in the same vector space (code+pre-trained models).

- Unified Evaluation Framework for Word Sense Disambiguation.

- BabelNet, a very large multilingual encyclopedic dictionary and semantic network.

- BabelDomains: Lexical items (synsets, Wikipedia pages) annotated with domains of knowledge.

- EuroSense: Multilingual sense annotations for Europarl.

- Supervised distributional framework (including Python API) for hypernym discovery.

- Find the word that does not belong: an evaluation benchmark for the outlier detection task

- Large multilingual corpus of sense-annotated textual definitions.

- Word similarity datasets in several languages (also cross-lingual!).


Looking forward to hearing from you!

invited talks / tutorials

Jose Camacho Collados

COLING 2020 Tutorial on Embeddings in NLP, 13 December 2020. [slides+video]

The Power of Natural Language Processing for Social Media Analysis, 13 October 2020, AI Tech North, UK. [slides]

Natural Language Processing at Cardiff University, 14 July 2020, Wales Tech Week, UK.

Word, Sense and Contextualized Embeddings: Vector Representations of Meaning in NLP, 6 February 2020, AI Wales, UK. [slides]

Word, Sense and Contextualized Embeddings: Vector Representations of Meaning in NLP, 8 January 2020, Universidad de Granada, Spain.

Word, Sense and Contextualized Embeddings: Vector Representations of Meaning in Natural Language Processing, 18 March 2019, Data and Knowledge Engineering Seminar, Cardiff University, UK. [slides]

NAACL 2018 Tutorial on The Interplay between Lexical Resources and NLP, 1 June 2018, New Orleans, USA. [slides][website]

Semantic (Vector) Representations of Word Senses, Concepts and Entities and their Applications, 20 April 2017, University of Cambridge, UK. [slides]

ACL 2016 Tutorial on Semantic Representation of Word Senses and Concepts, 7 August 2016, Berlin, Germany. [slides]

Semantic Representations of Word Senses, Concepts and Entities and their Applications, 19 October 2016, Pompeu Fabra University, Barcelona, Spain. [slides]

Computational Semantic Representation, 5 June 2015, Department of Applied Mathematics, University of California (UCLA), Los Angeles, USA.

Using NLP tools to identify people and places in a collection of 19th century emigrant letters, 'Digitising Experience of Migration' Workshop,
15 March 2014, Mellon Centre for Migration Studies, Omagh, Northern Ireland.