A5002307731- Topic Modeling
- Natural Language Processing Techniques
- Biomedical Text Mining and Ontologies
- Advanced Text Analysis Techniques
- Text Readability and Simplification
- Sentiment Analysis and Opinion Mining
- Semantic Web and Ontologies
- Advanced Graph Neural Networks
- Video Analysis and Summarization
- Software Testing and Debugging Techniques
- Speech and dialogue systems
- Machine Learning and Data Classification
- BIM and Construction Integration
- Knowledge Management and Technology
- Opinion Dynamics and Social Influence
- Data Quality and Management
- Occupational Health and Safety Research
- Machine Learning in Healthcare
- Software Engineering Research
- Mental Health via Writing
IT University of Copenhagen
2023
Tokyo Institute of Technology
2023
Administration for Community Living
2023
American Jewish Committee
2023
University of Manchester
2018-2022
Huawei Technologies (United Kingdom)
2022
National Institute of Advanced Industrial Science and Technology
2018-2019
Open Text (Canada)
2019
National Technical University of Athens
2016-2018
Inter-sentence relation extraction deals with a number of complex semantic relationships in documents, which require local, non-local, syntactic and dependencies. Existing methods do not fully exploit such We present novel inter-sentence model that builds labelled edge graph convolutional neural network on document-level graph. The is constructed using various inter- intra-sentence dependencies to capture local non-local dependency information. In order predict the an entity pair, we utilise...
Fenia Christopoulou, Makoto Miwa, Sophia Ananiadou. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint (EMNLP-IJCNLP). 2019.
We present a novel graph-based neural network model for relation extraction. Our treats multiple pairs in sentence simultaneously and considers interactions among them. All the entities are placed as nodes fully-connected graph structure. The edges represented with position-aware contexts around entity pairs. In order to consider different paths between two entities, we construct up l-length walks each pair. resulting merged iteratively used update edge representations into longer...
Identification of drugs, associated medication entities, and interactions among them are crucial to prevent unwanted effects drug therapy, known as adverse events. This article describes our participation the n2c2 shared-task in extracting relations between medication-related entities electronic health records.We proposed an ensemble approach for relation extraction classification drugs entities. We incorporated state-of-the-art named-entity recognition (NER) models based on bidirectional...
We present PanGu-Coder, a pretrained decoder-only language model adopting the PanGu-Alpha architecture for text-to-code generation, i.e. synthesis of programming solutions given natural problem description. train PanGu-Coder using two-stage strategy: first stage employs Causal Language Modelling (CLM) to pre-train on raw data, while second uses combination and Masked (MLM) training objectives that focus downstream task generation loosely curated pairs program definitions code functions....
Elisavet Palogiannidi, Athanasia Kolovou, Fenia Christopoulou, Filippos Kokkinos, Elias Iosif, Nikolaos Malandrakis, Haris Papageorgiou, Shrikanth Narayanan, Alexandros Potamianos. Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016). 2016.
Fenia Christopoulou, Makoto Miwa, Sophia Ananiadou. Proceedings of the 2021 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies. 2021.
Machine reading (MR) is essential for unlocking valuable knowledge contained in millions of existing biomedical documents. Over the last two decades1,2, most dramatic advances MR have followed wake critical corpus development3. Large, well-annotated corpora been associated with punctuated methodology and automated extraction systems same way that ImageNet4 was fundamental developing machine vision techniques. This study contributes six components to an advanced, named entity analysis tool...
Nested and overlapping events are particularly frequent informative structures in biomedical event extraction. However, state-of-the-art neural models either neglect those during learning or use syntactic features external tools to detect them. To overcome these limitations, this paper presents compares two models: a novel EXhaustive Neural Network (EXNN) Search-Based (SBNN) for detection of nested events.We evaluate the proposed as an component isolation within pipeline setting. Evaluation...
Accurate alignment between languages is fundamental for improving cross-lingual pre-trained language models (XLMs). Motivated by the natural phenomenon of code-switching (CS) in multilingual speakers, CS has been used as an effective data augmentation method that offers at word- or phrase-level, contrast to sentence-level via parallel instances. Existing approaches either use dictionaries sentences with word-alignment generate randomly switching words a sentence. However, such methods can be...
Typically, Distributional Semantic Models (DSMs) estimate semantic similarity between words using a single-model, where the multiple senses of polysemous are conflated in single representation. Similarly, textual affective analysis tasks, ambiguous usually not treated differently when estimating word scores. In this work, mixture model is proposed enabling combination scores estimated across topic-specific DSMs (TDSMs). Based on assumption that implies similarity, we extend to perform...
Large pre-trained language models have recently been expanded and applied to programming tasks with great success, often through further pre-training of a strictly-natural model--where training sequences typically contain both natural (linearised) language. Such approaches effectively map modalities the sequence into same embedding space. However, keywords (e.g. ``while'') very strictly defined semantics. As such, transfer learning from their usage may not necessarily be beneficial code...
Large language models (LLMs) have shown remarkable capabilities, but still struggle with processing extensive contexts, limiting their ability to maintain coherence and accuracy over long sequences. In contrast, the human brain excels at organising retrieving episodic experiences across vast temporal scales, spanning a lifetime. this work, we introduce EM-LLM, novel approach that integrates key aspects of memory event cognition into LLMs, enabling them effectively handle practically infinite...
Preference Optimization (PO) has proven an effective step for aligning language models to human-desired behaviors. Current variants, following the offline Direct objective, have focused on a strict setting where all tokens are contributing signals of KL divergence and rewards loss function. However, human preference is not affected by each word in sequence equally but often dependent specific words or phrases, e.g. existence toxic terms leads non-preferred responses. Based this observation,...
Curriculum Learning (CL) is a technique of training models via ranking examples in typically increasing difficulty trend with the aim accelerating convergence and improving generalisability. Current approaches for Natural Language Understanding (NLU) tasks use CL to improve in-distribution data performance often heuristic-oriented or task-agnostic difficulties. In this work, instead, we employ NLU by taking advantage dynamics as metrics, i.e., statistics that measure behavior model at hand...
Document-level relation extraction is a complex human process that requires logical inference to extract relationships between named entities in text. Existing approaches use graph-based neural models with words as nodes and edges relations them, encode across sentences. These are node-based, i.e., they form pair representations based solely on the two target node representations. However, entity can be better expressed through unique edge formed paths nodes. We thus propose an edge-oriented...
Machine reading is essential for unlocking valuable knowledge contained in the millions of existing biomedical documents. Over last two decades 1,2 , most dramatic advances machine-reading have followed wake critical corpus development 3 . Large, well-annotated corpora been associated with punctuated machine methodology and automated extraction systems same way that ImageNet 4 was fundamental developing vision techniques. This study contributes six components to an advanced, named-entity...
Inter-sentence relation extraction deals with a number of complex semantic relationships in documents, which require local, non-local, syntactic and dependencies. Existing methods do not fully exploit such We present novel inter-sentence model that builds labelled edge graph convolutional neural network on document-level graph. The is constructed using various inter- intra-sentence dependencies to capture local non-local dependency information. In order predict the an entity pair, we utilise...
We present a novel graph-based neural network model for relation extraction. Our treats multiple pairs in sentence simultaneously and considers interactions among them. All the entities are placed as nodes fully-connected graph structure. The edges represented with position-aware contexts around entity pairs. In order to consider different paths between two entities, we construct up l-length walks each pair. resulting merged iteratively used update edge representations into longer...