Our collective attention span is shortened by the flood of online information. With \textit{FarFetched}, we address the need for automated claim validation based on the aggregated evidence derived from multiple online news sources. We introduce an entity-centric reasoning framework in which latent connections between events, actions, or statements are revealed via entity mentions and represented in a graph database. Using entity linking and semantic similarity, we offer a way for collecting and combining information from diverse sources in order to generate evidence relevant to the user's claim. Then, we leverage textual entailment recognition to quantitatively determine whether this assertion is credible, based on the created evidence. Our approach tries to fill the gap in automated claim validation for less-resourced languages and is showcased on the Greek language, complemented by the training of relevant semantic textual similarity (STS) and natural language inference (NLI) models that are evaluated on translated versions of common benchmarks.
Cone beam computed tomography (CBCT) is a common way of diagnosing dental related diseases. Accurate segmentation of 3D tooth is of importance for the treatment. Although deep learning based methods have achieved convincing results in medical image processing, they need a large of annotated data for network training, making it very time-consuming in data collection and annotation. Besides, domain shift widely existing in the distribution of data acquired by different devices impacts severely the model generalization. To resolve the problem, we propose a multi-stage framework for 3D tooth segmentation in dental CBCT, which achieves the third place in the "Semi-supervised Teeth Segmentation" 3D (STS-3D) challenge. The experiments on validation set compared with other semi-supervised segmentation methods further indicate the validity of our approach.
Despite advancements in Natural Language Processing (NLP) and the growing availability of pretrained models, the English language remains the primary focus of model development. Continued pretraining on language-specific corpora provides a practical solution for adapting models to other languages. However, the impact of different pretraining settings on downstream tasks remains underexplored. This work introduces $\texttt{ptt5-v2}$, investigating the continued pretraining of T5 models for Portuguese. We first develop a baseline set of settings and pretrain models with sizes up to 3B parameters. Finetuning on three Portuguese downstream tasks (assin2 STS, assin2 RTE, and TweetSentBR) yields SOTA results on the latter two. We then explore the effects of different pretraining configurations, including pretraining data quality, optimization strategies, and multi-epoch pretraining. Perhaps surprisingly, their impact remains subtle compared to our baseline. We release $\texttt{ptt5-v2}$ pretrained checkpoints and their MonoT5-based finetuned $\texttt{MonoPTT5}$ rerankers on HuggingFace in their respective collections at \url{https://huggingface.co/unicamp-dl}.
Semantic Textual Similarity (STS) constitutes a critical research direction in computational linguistics and serves as a key indicator of the encoding capabilities of embedding models. Driven by advances in pre-trained language models and contrastive learning, leading sentence representation methods have reached an average Spearman's correlation score of approximately 86 across seven STS benchmarks in SentEval. However, further progress has become increasingly marginal, with no existing method attaining an average score higher than 86.5 on these tasks. This paper conducts an in-depth analysis of this phenomenon and concludes that the upper limit for Spearman's correlation scores under contrastive learning is 87.5. To transcend this ceiling, we propose an innovative approach termed Pcc-tuning, which employs Pearson's correlation coefficient as a loss function to refine model performance beyond contrastive learning. Experimental results demonstrate that Pcc-tuning can markedly surpass previous state-of-the-art strategies with only a minimal amount of fine-grained annotated samples.
Since the introduction of BERT and RoBERTa, research on Semantic Textual Similarity (STS) has made groundbreaking progress. Particularly, the adoption of contrastive learning has substantially elevated state-of-the-art performance across various STS benchmarks. However, contrastive learning categorizes text pairs as either semantically similar or dissimilar, failing to leverage fine-grained annotated information and necessitating large batch sizes to prevent model collapse. These constraints pose challenges for researchers engaged in STS tasks that involve nuanced similarity levels or those with limited computational resources, compelling them to explore alternatives like Sentence-BERT. Despite its efficiency, Sentence-BERT tackles STS tasks from a classification perspective, overlooking the progressive nature of semantic relationships, which results in suboptimal performance. To bridge this gap, this paper presents an innovative regression framework and proposes two simple yet effective loss functions: Translated ReLU and Smooth K2 Loss. Experimental results demonstrate that our method achieves convincing performance across seven established STS benchmarks and offers the potential for further optimization of contrastive learning pre-trained models.
One of the principal objectives of Natural Language Processing (NLP) is to generate meaningful representations from text. Improving the informativeness of the representations has led to a tremendous rise in the dimensionality and the memory footprint. It leads to a cascading effect amplifying the complexity of the downstream model by increasing its parameters. The available techniques cannot be applied to cross-modal applications such as text-to-image. To ameliorate these issues, a novel Text-to-Image methodology for generating fixed-length representations through a self-supervised Variational Auto-Encoder (VAE) for semantic evaluation applying transformers (TexIm FAST) has been proposed in this paper. The pictorial representations allow oblivious inference while retaining the linguistic intricacies, and are potent in cross-modal applications. TexIm FAST deals with variable-length sequences and generates fixed-length representations with over 75% reduced memory footprint. It enhances the efficiency of the models for downstream tasks by reducing its parameters. The efficacy of TexIm FAST has been extensively analyzed for the task of Semantic Textual Similarity (STS) upon the MSRPC, CNN/ Daily Mail, and XSum data-sets. The results demonstrate 6% improvement in accuracy compared to the baseline and showcase its exceptional ability to compare disparate length sequences such as a text with its summary.
Semantic textual similarity (STS) is a fundamental NLP task that measures the semantic similarity between a pair of sentences. In order to reduce the inherent ambiguity posed from the sentences, a recent work called Conditional STS (C-STS) has been proposed to measure the sentences' similarity conditioned on a certain aspect. Despite the popularity of C-STS, we find that the current C-STS dataset suffers from various issues that could impede proper evaluation on this task. In this paper, we reannotate the C-STS validation set and observe an annotator discrepancy on 55% of the instances resulting from the annotation errors in the original label, ill-defined conditions, and the lack of clarity in the task definition. After a thorough dataset analysis, we improve the C-STS task by leveraging the models' capability to understand the conditions under a QA task setting. With the generated answers, we present an automatic error identification pipeline that is able to identify annotation errors from the C-STS data with over 80% F1 score. We also propose a new method that largely improves the performance over baselines on the C-STS data by training the models with the answers. Finally we discuss the conditionality annotation based on the typed-feature structure (TFS) of entity types. We show in examples that the TFS is able to provide a linguistic foundation for constructing C-STS data with new conditions.
Speech-to-singing voice conversion (STS) task always suffers from data scarcity, because it requires paired speech and singing data. Compounding this issue are the challenges of content-pitch alignment and the suboptimal quality of generated outputs, presenting significant hurdles in STS research. This paper presents SVPT, an STS approach boosted by a self-supervised singing voice pre-training model. We leverage spoken language model techniques to tackle the rhythm alignment problem and the in-context learning capability to achieve zero-shot conversion. We adopt discrete-unit random resampling and pitch corruption strategies, enabling training with unpaired singing data and thus mitigating the issue of data scarcity. SVPT also serves as an effective backbone for singing voice synthesis (SVS), offering insights into scaling up SVS models. Experimental results indicate that SVPT delivers notable improvements in both STS and SVS endeavors. Audio samples are available at https://speech2sing.github.io.
Determining sentence pair similarity is crucial for various NLP tasks. A common technique to address this is typically evaluated on a continuous semantic textual similarity scale from 0 to 5. However, based on a linguistic observation in STS annotation guidelines, we found that the score in the range [4,5] indicates an upper-range sample, while the rest are lower-range samples. This necessitates a new approach to treating the upper-range and lower-range classes separately. In this paper, we introduce a novel embedding space decomposition method called MixSP utilizing a Mixture of Specialized Projectors, designed to distinguish and rank upper-range and lower-range samples accurately. The experimental results demonstrate that MixSP decreased the overlap representation between upper-range and lower-range classes significantly while outperforming competitors on STS and zero-shot benchmarks.
Most work treats large language models as black boxes without in-depth understanding of their internal working mechanism. In order to explain the internal representations of LLMs, we propose a gradient-based metric to assess the activation level of model parameters. Based on this metric, we obtain three preliminary findings. (1) When the inputs are in the same domain, parameters in the shallow layers will be activated densely, which means a larger portion of parameters will have great impacts on the outputs. In contrast, parameters in the deep layers are activated sparsely. (2) When the inputs are across different domains, parameters in shallow layers exhibit higher similarity in the activation behavior than deep layers. (3) In deep layers, the similarity of the distributions of activated parameters is positively correlated to the empirical data relevance. Further, we develop three validation experiments to solidify these findings. (1) Firstly, starting from the first finding, we attempt to configure different prune ratios for different layers, and find this method can benefit model pruning. (2) Secondly, we find that a pruned model based on one calibration set can better handle tasks related to the calibration task than those not related, which validate the second finding. (3) Thirdly, Based on the STS-B and SICK benchmark, we find that two sentences with consistent semantics tend to share similar parameter activation patterns in deep layers, which aligns with our third finding. Our work sheds light on the behavior of parameter activation in LLMs, and we hope these findings will have the potential to inspire more practical applications.
Making decent multi-lingual sentence representations is critical to achieve high performances in cross-lingual downstream tasks. In this work, we propose a novel method to align multi-lingual embeddings based on the similarity of sentences measured by a pre-trained mono-lingual embedding model. Given translation sentence pairs, we train a multi-lingual model in a way that the similarity between cross-lingual embeddings follows the similarity of sentences measured at the mono-lingual teacher model. Our method can be considered as contrastive learning with soft labels defined as the similarity between sentences. Our experimental results on five languages show that our contrastive loss with soft labels far outperforms conventional contrastive loss with hard labels in various benchmarks for bitext mining tasks and STS tasks. In addition, our method outperforms existing multi-lingual embeddings including LaBSE, for Tatoeba dataset. The code is available at https://github.com/YAI12xLinq-B/IMASCL
Brain folding patterns have been reported to carry clinically relevant information. The brain folds mainly during the last trimester of pregnancy, and the process might be durably disturbed by preterm birth. Yet little is known about preterm-specific patterns. In this work, we train a self-supervised model (SimCLR) on the UKBioBank cohort (21070 adults) to represent the right superior temporal sulcus (STS) region and apply it to sulci images of 374 babies from the dHCP database, containing preterms and full-terms, and acquired at 40 weeks post-menstrual age. We find a lower variability in the preterm embeddings, supported by the identification of a knob pattern, missing in the extremely preterm population.
Dense vector representations for sentences made significant progress in recent years as can be seen on sentence similarity tasks. Real-world phrase retrieval applications, on the other hand, still encounter challenges for effective use of dense representations. We show that when target phrases reside inside noisy context, representing the full sentence with a single dense vector, is not sufficient for effective phrase retrieval. We therefore look into the notion of representing multiple, sub-sentence, consecutive word spans, each with its own dense vector. We show that this technique is much more effective for phrase mining, yet requires considerable compute to obtain useful span representations. Accordingly, we make an argument for contextualized word/token embeddings that can be aggregated for arbitrary word spans while maintaining the span's semantic meaning. We introduce a modification to the common contrastive loss used for sentence embeddings that encourages word embeddings to have this property. To demonstrate the effect of this method we present a dataset based on the STS-B dataset with additional generated text, that requires finding the best matching paraphrase residing in a larger context and report the degree of similarity to the origin phrase. We demonstrate on this dataset, how our proposed method can achieve better results without significant increase to compute.
Random numbers are incredibly important in a variety of fields, and the need for their validation remains important for safety. A Quantum Random Number Generator (QRNG) can theoretically generate truly random numbers, however their quality still needs to be thoroughly validated. Generally, the task of validating random numbers has been delegated to different statistical tests such as the tests from the NIST Statistical Test Suite (STS), which are often slow and only perform one test at a time. Our work presents a deep learning model utilizing the Transformer architecture that 1) performs multiple NIST STS tests at once, and 2) runs much faster. This model outputs multi-label classification results on passing these statistical tests. We performed a thorough hyper-parameter optimization to converge on the best possible model and as a result, achieved a high degree of accuracy with a Macro F1-score of above 0.96. We also compared this model to a conventional deep learning method (Long Short Term Memory Recurrent Neural Networks) to quantify randomness and showed our model achieved similar performances while being much more efficient and scalable. The high performance and efficiency of this Transformer-based deep learning model showed that it can be a viable replacement for the NIST STS for validating random numbers.
In single-zone multi-node systems (SZMRSs), temperature controls rely on a single probe near the thermostat, resulting in temperature discrepancies that cause thermal discomfort and energy waste. Augmenting smart thermostats (STs) with per-room sensors has gained acceptance by major ST manufacturers. This paper leverages additional sensory information to empirically characterize the services provided by buildings, including thermal comfort, energy efficiency, and demand response (DR). Utilizing room-level time-series data from 1,000 houses, metadata from 110,000 houses across the United States, and data from two real-world testbeds, we examine the limitations of SZMNSs and explore the potential of remote sensors. We discovered that comfortable DR durations (CDRDs) for rooms are typically 70% longer or 40% shorter than for the room with the thermostat. When averaging, rooms at the control temperature's bounds are typically deviated around -3{\deg}F to 2.5{\deg}F from the average. Moreover, in 95% of houses, we identified rooms experiencing notably higher solar gains compared to the rest of the rooms, while 85% and 70% of houses demonstrated lower heat input and poor insulation, respectively. Lastly, it became evident that the consumption of cooling energy escalates with the increase in the number of sensors, whereas heating usage experiences fluctuations ranging from -19% to +25%. This study serves as a benchmark for assessing the thermal comfort and DR services in the existing housing stock, while also highlighting the energy efficiency impacts of sensing technologies. Our approach sets the stage for more granular, precise control strategies of SZMNSs.
Large language models (LLMs) are increasingly being integrated into search engines to provide natural language responses tailored to user queries. Customers and end-users are also becoming more dependent on these models for quick and easy purchase decisions. In this work, we investigate whether recommendations from LLMs can be manipulated to enhance a product's visibility. We demonstrate that adding a strategic text sequence (STS) -- a carefully crafted message -- to a product's information page can significantly increase its likelihood of being listed as the LLM's top recommendation. To understand the impact of STS, we use a catalog of fictitious coffee machines and analyze its effect on two target products: one that seldom appears in the LLM's recommendations and another that usually ranks second. We observe that the strategic text sequence significantly enhances the visibility of both products by increasing their chances of appearing as the top recommendation. This ability to manipulate LLM-generated search responses provides vendors with a considerable competitive advantage and has the potential to disrupt fair market competition. Just as search engine optimization (SEO) revolutionized how webpages are customized to rank higher in search engine results, influencing LLM recommendations could profoundly impact content optimization for AI-driven search services. Code for our experiments is available at https://github.com/aounon/llm-rank-optimizer.
Semantic textual similartiy (STS) and information retrieval tasks (IR) tasks have been the two major avenues to record the progress of embedding models in the past few years. Under the emerging Retrieval-augmented Generation (RAG) paradigm, we envision the need to evaluate next-level language understanding abilities of embedding models, and take a conscious look at the reasoning abilities stored in them. Addressing this, we pose the question: Can retrievers solve reasoning problems? By transforming reasoning tasks into retrieval tasks, we find that without specifically trained for reasoning-level language understanding, current state-of-the-art retriever models may still be far from being competent for playing the role of assisting LLMs, especially in reasoning-intensive tasks. Moreover, albeit trained to be aware of instructions, instruction-aware IR models are often better off without instructions in inference time for reasoning tasks, posing an overlooked retriever-LLM behavioral gap for the research community to align. However, recent decoder-based embedding models show great promise in narrowing the gap, highlighting the pathway for embedding models to achieve reasoning-level language understanding. We also show that, although current off-the-shelf re-ranker models fail on these tasks, injecting reasoning abilities into them through fine-tuning still appears easier than doing so to bi-encoders, and we are able to achieve state-of-the-art performance across all tasks by fine-tuning a reranking model. We release Reasoning as Retrieval Benchmark (RAR-b), a holistic suite of tasks and settings to evaluate the reasoning abilities stored in retriever models. RAR-b is available at https://github.com/gowitheflow-1998/RAR-b.
Knowledge graphs play a pivotal role in various applications, such as question-answering and fact-checking. Abstract Meaning Representation (AMR) represents text as knowledge graphs. Evaluating the quality of these graphs involves matching them structurally to each other and semantically to the source text. Existing AMR metrics are inefficient and struggle to capture semantic similarity. We also lack a systematic evaluation benchmark for assessing structural similarity between AMR graphs. To overcome these limitations, we introduce a novel AMR similarity metric, rematch, alongside a new evaluation for structural similarity called RARE. Among state-of-the-art metrics, rematch ranks second in structural similarity; and first in semantic similarity by 1--5 percentage points on the STS-B and SICK-R benchmarks. Rematch is also five times faster than the next most efficient metric.
Previous work on multimodal sentence embedding has proposed multimodal contrastive learning and achieved promising results. However, by taking the rest of the batch as negative samples without reviewing when forming contrastive pairs, those studies encountered many suspicious and noisy negative examples, significantly affecting the methods' overall performance. In this work, we propose KDMCSE (Knowledge Distillation Multimodal contrastive learning of Sentence Embeddings), a novel approach that enhances the discrimination and generalizability of multimodal representation and inherits the knowledge from the teacher model to learn the difference between positive and negative instances and via that, can detect noisy and wrong negative samples effectively before they are calculated in the contrastive objective. Furthermore, to overcome the limitation of modeling the variation within negative pairs, we introduce a new contrastive objective, AdapACSE (Adaptive Angular Margin Supervised Contrastive Learning for Multimodal sentence embeddings), that enhances the discriminative representation by strengthening the margin within the angular space while capturing varying semantics within the negative. Experimental results on widely used Semantic Textual Similarity (STS) benchmarks demonstrate the effectiveness of our approach.
Congenital heart defects (CHD) are the most prevalent birth defects in the United States and surgical outcomes vary considerably across the country. The outcomes of treatment for CHD differ for specific patient subgroups, with non-Hispanic Black and Hispanic populations experiencing higher rates of mortality and morbidity. A valid comparison of outcomes within racial/ethnic subgroups is difficult given large differences in case-mix and small subgroup sizes. We propose a causal inference framework for outcome assessment and leverage advances in transfer learning to incorporate data from both target and source populations to help estimate causal effects while accounting for different sources of risk factor and outcome differences across populations. Using the Society of Thoracic Surgeons' Congenital Heart Surgery Database (STS-CHSD), we focus on a national cohort of patients undergoing the Norwood operation from 2016-2022 to assess operative mortality and morbidity outcomes across U.S. geographic regions by race/ethnicity. We find racial and ethnic outcome differences after controlling for potential confounding factors. While geography does not have a causal effect on outcomes for non-Hispanic Caucasian patients, non-Hispanic Black patients experience wide variability in outcomes with estimated 30-day mortality ranging from 5.9% (standard error 2.2%) to 21.6% (4.4%) across U.S. regions.