Pretrained language models have achieved state-of-the-art performance when adapted to a downstream NLP task. However, theoretical analysis of these models is scarce and challenging since the pretraining and downstream tasks can be very different. We propose an analysis framework that links the pretraining and downstream tasks with an underlying latent variable generative model of text -- the downstream classifier must recover a function of the posterior distribution over the latent variables. We analyze head tuning (learning a classifier on top of the frozen pretrained model) and prompt tuning in this setting. The generative model in our analysis is either a Hidden Markov Model (HMM) or an HMM augmented with a latent memory component, motivated by long-term dependencies in natural language. We show that 1) under certain non-degeneracy conditions on the HMM, simple classification heads can solve the downstream task, 2) prompt tuning obtains downstream guarantees with weaker non-degeneracy conditions, and 3) our recovery guarantees for the memory-augmented HMM are stronger than for the vanilla HMM because task-relevant information is easier to recover from the long-term memory. Experiments on synthetically generated data from HMMs back our theoretical findings.
While multitask and transfer learning has shown to improve the performance of neural networks in limited data settings, they require pretraining of the model on large datasets beforehand. In this paper, we focus on improving the performance of weakly supervised sound event detection in low data and noisy settings simultaneously without requiring any pretraining task. To that extent, we propose a shared encoder architecture with sound event detection as a primary task and an additional secondary decoder for a self-supervised auxiliary task. We empirically evaluate the proposed framework for weakly supervised sound event detection on a remix dataset of the DCASE 2019 task 1 acoustic scene data with DCASE 2018 Task 2 sounds event data under 0, 10 and 20 dB SNR. To ensure we retain the localisation information of multiple sound events, we propose a two-step attention pooling mechanism that provides a time-frequency localisation of multiple audio events in the clip. The proposed framework with two-step attention outperforms existing benchmark models by 22.3%, 12.8%, 5.9% on 0, 10 and 20 dB SNR respectively. We carry out an ablation study to determine the contribution of the auxiliary task and two-step attention pooling to the SED performance improvement.
Sexism has become an increasingly major problem on social networks during the last years. The first shared task on sEXism Identification in Social neTworks (EXIST) at IberLEF 2021 is an international competition in the field of Natural Language Processing (NLP) with the aim to automatically identify sexism in social media content by applying machine learning methods. Thereby sexism detection is formulated as a coarse (binary) classification problem and a fine-grained classification task that distinguishes multiple types of sexist content (e.g., dominance, stereotyping, and objectification). This paper presents the contribution of the AIT_FHSTP team at the EXIST2021 benchmark for both tasks. To solve the tasks we applied two multilingual transformer models, one based on multilingual BERT and one based on XLM-R. Our approach uses two different strategies to adapt the transformers to the detection of sexist content: first, unsupervised pre-training with additional data and second, supervised fine-tuning with additional and augmented data. For both tasks our best model is XLM-R with unsupervised pre-training on the EXIST data and additional datasets and fine-tuning on the provided dataset. The best run for the binary classification (task 1) achieves a macro F1-score of 0.7752 and scores 5th rank in the benchmark; for the multiclass classification (task 2) our best submission scores 6th rank with a macro F1-score of 0.5589.
We present the task description and discussion on the results of the DCASE 2021 Challenge Task 2. In 2020, we organized an unsupervised anomalous sound detection (ASD) task, identifying whether a given sound was normal or anomalous without anomalous training data. In 2021, we organized an advanced unsupervised ASD task under domain-shift conditions, which focuses on the inevitable problem of the practical use of ASD systems. The main challenge of this task is to detect unknown anomalous sounds where the acoustic characteristics of the training and testing samples are different, i.e., domain-shifted. This problem frequently occurs due to changes in seasons, manufactured products, and/or environmental noise. We received 75 submissions from 26 teams, and several novel approaches have been developed in this challenge. On the basis of the analysis of the evaluation results, we found that there are two types of remarkable approaches that TOP-5 winning teams adopted: 1) ensemble approaches of ``outlier exposure'' (OE)-based detectors and ``inlier modeling'' (IM)-based detectors and 2) approaches based on IM-based detection for features learned in a machine-identification task.
This paper describes the performance of the team cs60075_team2 at SemEval 2021 Task 1 - Lexical Complexity Prediction. The main contribution of this paper is to fine-tune transformer-based language models pre-trained on several text corpora, some being general (E.g., Wikipedia, BooksCorpus), some being the corpora from which the CompLex Dataset was extracted, and others being from other specific domains such as Finance, Law, etc. We perform ablation studies on selecting the transformer models and how their individual complexity scores are aggregated to get the resulting complexity scores. Our method achieves a best Pearson Correlation of $0.784$ in sub-task 1 (single word) and $0.836$ in sub-task 2 (multiple word expressions).
This paper describes the experimental framework and results of the ICDAR 2021 Competition on On-Line Signature Verification (SVC 2021). The goal of SVC 2021 is to evaluate the limits of on-line signature verification systems on popular scenarios (office/mobile) and writing inputs (stylus/finger) through large-scale public databases. Three different tasks are considered in the competition, simulating realistic scenarios as both random and skilled forgeries are simultaneously considered on each task. The results obtained in SVC 2021 prove the high potential of deep learning methods. In particular, the best on-line signature verification system of SVC 2021 obtained Equal Error Rate (EER) values of 3.33% (Task 1), 7.41% (Task 2), and 6.04% (Task 3). SVC 2021 will be established as an on-going competition, where researchers can easily benchmark their systems against the state of the art in an open common platform using large-scale public databases such as DeepSignDB and SVC2021_EvalDB, and standard experimental protocols.
This paper presents the results and main findings of SemEval-2021 Task 1 - Lexical Complexity Prediction. We provided participants with an augmented version of the CompLex Corpus (Shardlow et al 2020). CompLex is an English multi-domain corpus in which words and multi-word expressions (MWEs) were annotated with respect to their complexity using a five point Likert scale. SemEval-2021 Task 1 featured two Sub-tasks: Sub-task 1 focused on single words and Sub-task 2 focused on MWEs. The competition attracted 198 teams in total, of which 54 teams submitted official runs on the test data to Sub-task 1 and 37 to Sub-task 2.
This paper presents the final results of the ICDAR 2021 Competition on Historical Map Segmentation (MapSeg), encouraging research on a series of historical atlases of Paris, France, drawn at 1/5000 scale between 1894 and 1937. The competition featured three tasks, awarded separately. Task~1 consists in detecting building blocks and was won by the L3IRIS team using a DenseNet-121 network trained in a weakly supervised fashion. This task is evaluated on 3 large images containing hundreds of shapes to detect. Task~2 consists in segmenting map content from the larger map sheet, and was won by the UWB team using a U-Net-like FCN combined with a binarization method to increase detection edge accuracy. Task~3 consists in locating intersection points of geo-referencing lines, and was also won by the UWB team who used a dedicated pipeline combining binarization, line detection with Hough transform, candidate filtering, and template matching for intersection refinement. Tasks~2 and~3 are evaluated on 95 map sheets with complex content. Dataset, evaluation tools and results are available under permissive licensing at \url{https://icdar21-mapseg.github.io/}.
In this article, we present a description of our systems as a part of our participation in the shared task namely Artificial Intelligence for Legal Assistance (AILA 2019). This is an integral event of Forum for Information Retrieval Evaluation-2019. The outcomes of this track would be helpful for the automation of the working process of the Indian Judiciary System. The manual working procedures and documentation at any level (from lower to higher court) of the judiciary system are very complex in nature. The systems produced as a part of this track would assist the law practitioners. It would be helpful for common men too. This kind of track also opens the path of research of Natural Language Processing (NLP) in the judicial domain. This track defined two problems such as Task 1: Identifying relevant prior cases for a given situation and Task 2: Identifying the most relevant statutes for a given situation. We tackled both of them. Our proposed approaches are based on BM25 and Doc2Vec. As per the results declared by the task organizers, we are in 3rd and a modest position in Task 1 and Task 2 respectively.
Educational content labeled with proper knowledge components (KCs) are particularly useful to teachers or content organizers. However, manually labeling educational content is labor intensive and error-prone. To address this challenge, prior research proposed machine learning based solutions to auto-label educational content with limited success. In this work, we significantly improve prior research by (1) expanding the input types to include KC descriptions, instructional video titles, and problem descriptions (i.e., three types of prediction task), (2) doubling the granularity of the prediction from 198 to 385 KC labels (i.e., more practical setting but much harder multinomial classification problem), (3) improving the prediction accuracies by 0.5-2.3% using Task-adaptive Pre-trained BERT, outperforming six baselines, and (4) proposing a simple evaluation measure by which we can recover 56-73% of mispredicted KC labels. All codes and data sets in the experiments are available at:https://github.com/tbs17/TAPT-BERT
In this paper, we introduce MIMII DUE, a new dataset for malfunctioning industrial machine investigation and inspection with domain shifts due to changes in operational and environmental conditions. Conventional methods for anomalous sound detection face practical challenges because the distribution of features changes between the training and operational phases (called domain shift) due to various real-world factors. To check the robustness against domain shifts, we need a dataset that actually includes domain shifts, but such a dataset does not exist so far. The new dataset we created consists of the normal and abnormal operating sounds of five different types of industrial machines under two different operational/environmental conditions (source domain and target domain) independent of normal/abnormal, with domain shifts occurring between the two domains. Experimental results showed significant performance differences between the source and target domains, indicating that the dataset contains the domain shifts. These findings demonstrate that the dataset will be helpful for checking the robustness against domain shifts. The dataset is a subset of the dataset for DCASE 2021 Challenge Task 2 and freely available for download at https://zenodo.org/record/4740355
While deep reinforcement learning (RL) promises freedom from hand-labeled data, great successes, especially for Embodied AI, require significant work to create supervision via carefully shaped rewards. Indeed, without shaped rewards, i.e., with only terminal rewards, present-day Embodied AI results degrade significantly across Embodied AI problems from single-agent Habitat-based PointGoal Navigation (SPL drops from 55 to 0) and two-agent AI2-THOR-based Furniture Moving (success drops from 58% to 1%) to three-agent Google Football-based 3 vs. 1 with Keeper (game score drops from 0.6 to 0.1). As training from shaped rewards doesn't scale to more realistic tasks, the community needs to improve the success of training with terminal rewards. For this we propose GridToPix: 1) train agents with terminal rewards in gridworlds that generically mirror Embodied AI environments, i.e., they are independent of the task; 2) distill the learned policy into agents that reside in complex visual worlds. Despite learning from only terminal rewards with identical models and RL algorithms, GridToPix significantly improves results across tasks: from PointGoal Navigation (SPL improves from 0 to 64) and Furniture Moving (success improves from 1% to 25%) to football gameplay (game score improves from 0.1 to 0.6). GridToPix even helps to improve the results of shaped reward training.
This paper describes the Academia Sinica systems for the two tasks of Voice Conversion Challenge 2020, namely voice conversion within the same language (Task 1) and cross-lingual voice conversion (Task 2). For both tasks, we followed the cascaded ASR+TTS structure, using phonetic tokens as the TTS input instead of the text or characters. For Task 1, we used the international phonetic alphabet (IPA) as the input of the TTS model. For Task 2, we used unsupervised phonetic symbols extracted by the vector-quantized variational autoencoder (VQVAE). In the evaluation, the listening test showed that our systems performed well in the VCC2020 challenge.
This paper presents the PALI team's winning system for SemEval-2021 Task 2: Multilingual and Cross-lingual Word-in-Context Disambiguation. We fine-tune XLM-RoBERTa model to solve the task of word in context disambiguation, i.e., to determine whether the target word in the two contexts contains the same meaning or not. In the implementation, we first specifically design an input tag to emphasize the target word in the contexts. Second, we construct a new vector on the fine-tuned embeddings from XLM-RoBERTa and feed it to a fully-connected network to output the probability of whether the target word in the context has the same meaning or not. The new vector is attained by concatenating the embedding of the [CLS] token and the embeddings of the target word in the contexts. In training, we explore several tricks, such as the Ranger optimizer, data augmentation, and adversarial training, to improve the model prediction. Consequently, we attain first place in all four cross-lingual tasks.
This paper presents a preliminary experimentation study using the CLEF 2017 eHealth Task 2 collection for evaluating the effectiveness of different indexing methodologies of documents and query parsing techniques. Furthermore, it is an attempt to advance and share the efforts of observing the characteristics and helpfulness of various methodologies for indexing PubMed documents and for different topic parsing techniques to produce queries. For this purpose, my research includes experimentation with different document indexing methodologies, by utilising existing tools, such as the Lucene4IR (L4IR) information retrieval system, the Technology Assisted Reviews for Empirical Medicine tool for parsing topics of the CLEF collection and the TREC evaluation tool to appraise system's performance. The results showed that including a greater number of fields to the PubMed indexer of L4IR is a decisive factor for the retrieval effectiveness of L4IR.
This paper presents a set of experiments to evaluate and compare between the performance of using CBOW Word2Vec and Lemma2Vec models for Arabic Word-in-Context (WiC) disambiguation without using sense inventories or sense embeddings. As part of the SemEval-2021 Shared Task 2 on WiC disambiguation, we used the dev.ar-ar dataset (2k sentence pairs) to decide whether two words in a given sentence pair carry the same meaning. We used two Word2Vec models: Wiki-CBOW, a pre-trained model on Arabic Wikipedia, and another model we trained on large Arabic corpora of about 3 billion tokens. Two Lemma2Vec models was also constructed based on the two Word2Vec models. Each of the four models was then used in the WiC disambiguation task, and then evaluated on the SemEval-2021 test.ar-ar dataset. At the end, we reported the performance of different models and compared between using lemma-based and word-based models.
This paper presents our contribution to SemEval-2021 Task 2: Multilingual and Cross-lingual Word-in-Context Disambiguation (MCL-WiC). Our experiments cover English (EN-EN) sub-track from the multilingual setting of the task. We experiment with several pre-trained language models and investigate an impact of different top-layers on fine-tuning. We find the combination of Cosine Similarity and ReLU activation leading to the most effective fine-tuning procedure. Our best model results in accuracy 92.7%, which is the fourth-best score in EN-EN sub-track.
Identifying whether a word carries the same meaning or different meaning in two contexts is an important research area in natural language processing which plays a significant role in many applications such as question answering, document summarisation, information retrieval and information extraction. Most of the previous work in this area rely on language-specific resources making it difficult to generalise across languages. Considering this limitation, our approach to SemEval-2021 Task 2 is based only on pretrained transformer models and does not use any language-specific processing and resources. Despite that, our best model achieves 0.90 accuracy for English-English subtask which is very compatible compared to the best result of the subtask; 0.93 accuracy. Our approach also achieves satisfactory results in other monolingual and cross-lingual language pairs as well.
We describe the Uppsala NLP submission to SemEval-2021 Task 2 on multilingual and cross-lingual word-in-context disambiguation. We explore the usefulness of three pre-trained multilingual language models, XLM-RoBERTa (XLMR), Multilingual BERT (mBERT) and multilingual distilled BERT (mDistilBERT). We compare these three models in two setups, fine-tuning and as feature extractors. In the second case we also experiment with using dependency-based information. We find that fine-tuning is better than feature extraction. XLMR performs better than mBERT in the cross-lingual setting both with fine-tuning and feature extraction, whereas these two models give a similar performance in the multilingual setting. mDistilBERT performs poorly with fine-tuning but gives similar results to the other models when used as a feature extractor. We submitted our two best systems, fine-tuned with XLMR and mBERT.
This paper describes the IDLab submission for the text-independent task of the Short-duration Speaker Verification Challenge 2021 (SdSVC-21). This speaker verification competition focuses on short duration test recordings and cross-lingual trials, along with the constraint of limited availability of in-domain DeepMine Farsi training data. Currently, both Time Delay Neural Networks (TDNNs) and ResNets achieve state-of-the-art results in speaker verification. These architectures are structurally very different and the construction of hybrid networks looks a promising way forward. We introduce a 2D convolutional stem in a strong ECAPA-TDNN baseline to transfer some of the strong characteristics of a ResNet based model to this hybrid CNN-TDNN architecture. Similarly, we incorporate absolute frequency positional encodings in an SE-ResNet34 architecture. These learnable feature map biases along the frequency axis offer this architecture a straightforward way to exploit frequency positional information. We also propose a frequency-wise variant of Squeeze-Excitation (SE) which better preserves frequency-specific information when rescaling the feature maps. Both modified architectures significantly outperform their corresponding baseline on the SdSVC-21 evaluation data and the original VoxCeleb1 test set. A four system fusion containing the two improved architectures achieved a third place in the final SdSVC-21 Task 2 ranking.