Antoine J.‐P. Tixier

Christos Xypolopoulos, Antoine J. -P. Tixier, Michalis Vazirgiannis

The number of senses of a given word, or polysemy, is a very subjective notion, which varies widely across annotators and resources. We propose a novel method to estimate polysemy, based on simple geometry in the contextual embedding space. Our approach is fully unsupervised and purely data-driven. We show through rigorous experiments that our rankings are well correlated (with strong statistical significance) with 6 different rankings derived from famous human-constructed resources such as WordNet, OntoNotes, Oxford, Wikipedia etc., for 6 different standard metrics. We also visualize and analyze the correlation between the human rankings. A valuable by-product of our method is the ability to sample, at no extra cost, sentences containing different senses of a given word. Finally, the fully unsupervised nature of our method makes it applicable to any language. Code and data are publicly available at https://github.com/ksipos/polysemy-assessment . The paper was accepted as a long paper at EACL 2021.

Giannis Nikolentzos, Antoine J. -P. Tixier, Michalis Vazirgiannis

Graph neural networks have recently emerged as a very effective framework for processing graph-structured data. These models have achieved state-of-the-art performance in many tasks. Most graph neural networks can be described in terms of message passing, vertex update, and readout functions. In this paper, we represent documents as word co-occurrence networks and propose an application of the message passing framework to NLP, the Message Passing Attention network for Document understanding (MPAD). We also propose several hierarchical variants of MPAD. Experiments conducted on 10 standard text classification datasets show that our architectures are competitive with the state-of-the-art. Ablation studies reveal further insights about the impact of the different components on performance. Code is publicly available at: https://github.com/giannisnik/mpad .

Moussa Kamal Eddine, Guokan Shang, Antoine J. -P. Tixier et al.

Fast and reliable evaluation metrics are key to R&D progress. While traditional natural language generation metrics are fast, they are not very reliable. Conversely, new metrics based on large pretrained language models are much more reliable, but require significant computational resources. In this paper, we propose FrugalScore, an approach to learn a fixed, low cost version of any expensive NLG metric, while retaining most of its original performance. Experiments with BERTScore and MoverScore on summarization and translation show that FrugalScore is on par with the original metrics (and sometimes better), while having several orders of magnitude less parameters and running several times faster. On average over all learned metrics, tasks, and variants, FrugalScore retains 96.8% of the performance, runs 24 times faster, and has 35 times less parameters than the original metrics. We make our trained metrics publicly available, to benefit the entire NLP community and in particular researchers and practitioners with limited resources.

Moussa Kamal Eddine, Antoine J. -P. Tixier, Michalis Vazirgiannis

Inductive transfer learning has taken the entire NLP field by storm, with models such as BERT and BART setting new state of the art on countless NLU tasks. However, most of the available models and research have been conducted for English. In this work, we introduce BARThez, the first large-scale pretrained seq2seq model for French. Being based on BART, BARThez is particularly well-suited for generative tasks. We evaluate BARThez on five discriminative tasks from the FLUE benchmark and two generative tasks from a novel summarization dataset, OrangeSum, that we created for this research. We show BARThez to be very competitive with state-of-the-art BERT-based French language models such as CamemBERT and FlauBERT. We also continue the pretraining of a multilingual BART on BARThez' corpus, and show our resulting model, mBARThez, to significantly boost BARThez' generative performance. Code, data and models are publicly available.

Guokan Shang, Antoine Jean-Pierre Tixier, Michalis Vazirgiannis et al.

Recent work in Dialogue Act (DA) classification approaches the task as a sequence labeling problem, using neural network models coupled with a Conditional Random Field (CRF) as the last layer. CRF models the conditional probability of the target DA label sequence given the input utterance sequence. However, the task involves another important input sequence, that of speakers, which is ignored by previous work. To address this limitation, this paper proposes a simple modification of the CRF layer that takes speaker-change into account. Experiments on the SwDA corpus show that our modified CRF layer outperforms the original one, with very wide margins for some DA labels. Further, visualizations demonstrate that our CRF layer can learn meaningful, sophisticated transition patterns between DA label pairs conditioned on speaker-change in an end-to-end way. Code is publicly available.

Jean-Baptiste Remy, Antoine Jean-Pierre Tixier, Michalis Vazirgiannis

The Hierarchical Attention Network (HAN) has made great strides, but it suffers a major limitation: at level 1, each sentence is encoded in complete isolation. In this work, we propose and compare several modifications of HAN in which the sentence encoder is able to make context-aware attentional decisions (CAHAN). Furthermore, we propose a bidirectional document encoder that processes the document forwards and backwards, using the preceding and following sentences as context. Experiments on three large-scale sentiment and topic classification datasets show that the bidirectional version of CAHAN outperforms HAN everywhere, with only a modest increase in computation time. While results are promising, we expect the superiority of CAHAN to be even more evident on tasks requiring a deeper understanding of the input documents, such as abstractive summarization. Code is publicly available.

Henrietta Baker, Matthew R. Hallowell, Antoine J. -P. Tixier

In light of the increasing availability of digitally recorded safety reports in the construction industry, it is important to develop methods to exploit these data to improve our understanding of safety incidents and ability to learn from them. In this study, we compare several approaches to automatically learn injury precursors from raw construction accident reports. More precisely, we experiment with two state-of-the-art deep learning architectures for Natural Language Processing (NLP), Convolutional Neural Networks (CNN) and Hierarchical Attention Networks (HAN), and with the established Term Frequency - Inverse Document Frequency representation (TF-IDF) + Support Vector Machine (SVM) approach. For each model, we provide a method to identify (after training) the textual patterns that are, on average, the most predictive of each safety outcome. We show that among those pieces of text, valid injury precursors can be found. The proposed methods can also be used by the user to visualize and understand the models' predictions.

Guokan Shang, Antoine Jean-Pierre Tixier, Michalis Vazirgiannis et al.

Abstractive community detection is an important spoken language understanding task, whose goal is to group utterances in a conversation according to whether they can be jointly summarized by a common abstractive sentence. This paper provides a novel approach to this task. We first introduce a neural contextual utterance encoder featuring three types of self-attention mechanisms. We then train it using the siamese and triplet energy-based meta-architectures. Experiments on the AMI corpus show that our system outperforms multiple energy-based and non-energy based baselines from the state-of-the-art. Code and data are publicly available.

Antoine J. -P. Tixier

My notes on Deep Learning for NLP.

Antoine J. -P. Tixier, Maria-Evgenia G. Rossi, Fragkiskos D. Malliaros et al.

Some of the most effective influential spreader detection algorithms are unstable to small perturbations of the network structure. Inspired by bagging in Machine Learning, we propose the first Perturb and Combine (P&C) procedure for networks. It (1) creates many perturbed versions of a given graph, (2) applies a node scoring function separately to each graph, and (3) combines the results. Experiments conducted on real-world networks of various sizes with the k-core, generalized k-core, and PageRank algorithms reveal that P&C brings substantial improvements. Moreover, this performance boost can be obtained at almost no extra cost through parallelization. Finally, a bias-variance analysis suggests that P&C works mainly by reducing bias, and that therefore, it should be capable of improving the performance of all vertex scoring functions, including stable ones.

Guokan Shang, Wensi Ding, Zekun Zhang et al.

We introduce a novel graph-based framework for abstractive meeting speech summarization that is fully unsupervised and does not rely on any annotations. Our work combines the strengths of multiple recent approaches while addressing their weaknesses. Moreover, we leverage recent advances in word embeddings and graph degeneracy applied to NLP to take exterior semantic knowledge into account, and to design custom diversity and informativeness measures. Experiments on the AMI and ICSI corpus show that our system improves on the state-of-the-art. Code and data are publicly available, and our system can be interactively tested.

Giannis Nikolentzos, Polykarpos Meladianos, Antoine Jean-Pierre Tixier et al.

Graph kernels have been successfully applied to many graph classification problems. Typically, a kernel is first designed, and then an SVM classifier is trained based on the features defined implicitly by this kernel. This two-stage approach decouples data representation from learning, which is suboptimal. On the other hand, Convolutional Neural Networks (CNNs) have the capability to learn their own features directly from the raw data during training. Unfortunately, they cannot handle irregular data such as graphs. We address this challenge by using graph kernels to embed meaningful local neighborhoods of the graphs in a continuous vector space. A set of filters is then convolved with these patches, pooled, and the output is then passed to a feedforward network. With limited parameter tuning, our approach outperforms strong baselines on 7 out of 10 benchmark datasets.

Antoine Jean-Pierre Tixier, Giannis Nikolentzos, Polykarpos Meladianos et al.

Graph learning is currently dominated by graph kernels, which, while powerful, suffer some significant limitations. Convolutional Neural Networks (CNNs) offer a very appealing alternative, but processing graphs with CNNs is not trivial. To address this challenge, many sophisticated extensions of CNNs have recently been introduced. In this paper, we reverse the problem: rather than proposing yet another graph CNN model, we introduce a novel way to represent graphs as multi-channel image-like structures that allows them to be handled by vanilla 2D CNNs. Experiments reveal that our method is more accurate than state-of-the-art graph kernels and graph CNNs on 4 out of 6 real-world datasets (with and without continuous node attributes), and close elsewhere. Our approach is also preferable to graph kernels in terms of time complexity. Code and data are publicly available.

Antoine J. -P. Tixier, Michalis Vazirgiannis, Matthew R. Hallowell

We introduce word vectors for the construction domain. Our vectors were obtained by running word2vec on an 11M-word corpus that we created from scratch by leveraging freely-accessible online sources of construction-related text. We first explore the embedding space and show that our vectors capture meaningful construction-specific concepts. We then evaluate the performance of our vectors against that of ones trained on a 100B-word corpus (Google News) within the framework of an injury report classification task. Without any parameter tuning, our embeddings give competitive results, and outperform the Google News vectors in many cases. Using a keyword-based compression of the reports also leads to a significant speed-up with only a limited loss in performance. We release our corpus and the data set we created for the classification task as publicly available, in the hope that they will be used by future studies for benchmarking and building on our work.

Antoine J. -P. Tixier, Matthew R. Hallowell, Balaji Rajagopalan

By building on a recently introduced genetic-inspired attribute-based conceptual framework for safety risk analysis, we propose a novel methodology to compute construction univariate and bivariate construction safety risk at a situational level. Our fully data-driven approach provides construction practitioners and academicians with an easy and automated way of extracting valuable empirical insights from databases of unstructured textual injury reports. By applying our methodology on an attribute and outcome dataset directly obtained from 814 injury reports, we show that the frequency-magnitude distribution of construction safety risk is very similar to that of natural phenomena such as precipitation or earthquakes. Motivated by this observation, and drawing on state-of-the-art techniques in hydroclimatology and insurance, we introduce univariate and bivariate nonparametric stochastic safety risk generators, based on Kernel Density Estimators and Copulas. These generators enable the user to produce large numbers of synthetic safety risk values faithfully to the original data, allowing safetyrelated decision-making under uncertainty to be grounded on extensive empirical evidence. Just like the accurate modeling and simulation of natural phenomena such as wind or streamflow is indispensable to successful structure dimensioning or water reservoir management, we posit that improving construction safety calls for the accurate modeling, simulation, and assessment of safety risk. The underlying assumption is that like natural phenomena, construction safety may benefit from being studied in an empirical and quantitative way rather than qualitatively which is the current industry standard. Finally, a side but interesting finding is that attributes related to high energy levels and to human error emerge as strong risk shapers on the dataset we used to illustrate our methodology.