Cross domain similarities and intra-person changes

Add to your list(s) Download to your calendar using vCal

• Maria Liakata (University of Warwick)
• Friday 30 April 2021, 12:00-13:00
• Virtual (Zoom).

If you have a question about this talk, please contact Huiyuan Xie.

Join Zoom Meeting https://cl-cam-ac-uk.zoom.us/j/97474773044?pwd=ZzZsQTZPVFdRWnFUNElxS3RlQzRXdz09

Meeting ID: 974 7477 3044 Passcode: 322272

I will talk about two conceptually interconnected lines of work in NLP within my group; on the one hand identifying semantic similarities between instances (sentences or longer texts but also entities) across domains, and on the other hand detecting changes within the same person or domain over time. Even though semantic similarity is a fundamental task within NLP it can be very challenging when comparisons are made across domains as the vocabulary and context can be very different from one domain setting to another. I will talk about recent work of ours where we address semantic similarity between two texts in a variety of datasets, including community question answering, by injecting domain-specific topic model information to pre-trained language models [1]. I will also be discussing how in the case of cross domain entity similarity (and co-reference more specifically) current models struggle, some of the reasons behind this and a new resource to help with addressing this problem [2]. The second part of my talk can be seen as the flip side of semantic similarity, where the goal is to look for differences in the representation of the same individual (word or person) that signal a change. I will be discussing work of ours on sequential modelling of the evolution of a word for semantic change detection [3] and how we are developing methods to detect changes in individuals as part of my UKRI Turing AI fellowship.

[1] Peinelt, N., Nguyen, D., & Liakata, M. (2020, July). tBERT: Topic models and BERT joining forces for semantic similarity detection. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (pp. 7047-7055). https://www.aclweb.org/anthology/2020.acl-main.630/ [2] Ravenscroft, J., Clare, A., Cattan, A., Dagan, I., & Liakata, M. (2021, April). CDˆ2CR: Co-reference resolution across documents and domains. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume (pp. 270-280). https://www.aclweb.org/anthology/2021.eacl-main.21/ [3] Tsakalidis, A., & Liakata, M. (2020, November). Sequential Modelling of the Evolution of Word Representations for Semantic Change Detection. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) (pp. 8485-8497). https://www.aclweb.org/anthology/2020.emnlp-main.682.pdf

This talk is part of the NLIP Seminar Series series.

This talk is included in these lists:

• All Talks (aka the CURE list)
• Cambridge Centre for Data-Driven Discovery (C2D3)
• Cambridge Forum of Science and Humanities
• Cambridge Language Sciences
• Cambridge talks
• Chris Davis' list
• Department of Computer Science and Technology talks and seminars
• Graduate-Seminars
• Guy Emerson's list
• Interested Talks
• Language Sciences for Graduate Students
• NLIP Seminar Series
• PMRFPS's
• Simon Baker's List
• Trust & Technology Initiative - interesting events
• Virtual (Zoom)
• bld31
• ndk22's list
• ob366-ai4er
• rp587

Note that ex-directory lists are not shown.