Extraction of concepts and entities of interest from non-formal texts such as social media posts and informal communication is an important capability for decision support systems in many domains, including healthcare, customer relationship management, and others. Despite the recent advances in training large language models for a variety of natural language processing tasks, the developed models and techniques have mainly focused on formal texts and do not perform as well on colloquial data, which is characterized by a number of distinct challenges. In our research, we focus on the healthcare domain and investigate the problem of symptom recognition from colloquial texts by designing and evaluating several training strategies for BERT-based model fine-tuning. These strategies are distinguished by the choice of the base model, the training corpora, and application of term perturbations in the training data. The best-performing models trained using these strategies outperform the state-of-the-art specialized symptom recognizer by a large margin. Through a series of experiments, we have found specific patterns of model behavior associated with the training strategies we designed. We present design principles for training strategies for effective entity recognition in colloquial texts based on our findings.
Mapping speech tokens to the same feature space as text tokens has become the paradigm for the integration of speech modality into decoder-only large language models (LLMs). An alternative approach is to use an encoder-decoder architecture that incorporates speech features through cross-attention. This approach, however, has received less attention in the literature. In this work, we connect the Whisper encoder with ChatGLM3 and provide in-depth comparisons of these two approaches using Chinese automatic speech recognition (ASR) and name entity recognition (NER) tasks. We evaluate them not only by conventional metrics like the F1 score but also by a novel fine-grained taxonomy of ASR-NER errors. Our experiments reveal that encoder-decoder architecture outperforms decoder-only architecture with a short context, while decoder-only architecture benefits from a long context as it fully exploits all layers of the LLM. By using LLM, we significantly reduced the entity omission errors and improved the entity ASR accuracy compared to the Conformer baseline. Additionally, we obtained a state-of-the-art (SOTA) F1 score of 0.805 on the AISHELL-NER test set by using chain-of-thought (CoT) NER which first infers long-form ASR transcriptions and then predicts NER labels.
Diffusion models have demonstrated promising results in text-to-audio generation tasks. However, their practical usability is hindered by slow sampling speeds, limiting their applicability in high-throughput scenarios. To address this challenge, progressive distillation methods have been effective in producing more compact and efficient models. Nevertheless, these methods encounter issues with unbalanced weights at both high and low noise levels, potentially impacting the quality of generated samples. In this paper, we propose the adaptation of the progressive distillation method to text-to-audio generation tasks and introduce the Balanced SNR-Aware~(BSA) method, an enhanced loss-weighting mechanism for diffusion distillation. The BSA method employs a balanced approach to weight the loss for both high and low noise levels. We evaluate our proposed method on the AudioCaps dataset and report experimental results showing superior performance during the reverse diffusion process compared to previous distillation methods with the same number of sampling steps. Furthermore, the BSA method allows for a significant reduction in sampling steps from 200 to 25, with minimal performance degradation when compared to the original teacher models.
Named Entity Recognition seeks to extract substrings within a text that name real-world objects and to determine their type (for example, whether they refer to persons or organizations). In this survey, we first present an overview of recent popular approaches, but we also look at graph- and transformer- based methods including Large Language Models (LLMs) that have not had much coverage in other surveys. Second, we focus on methods designed for datasets with scarce annotations. Third, we evaluate the performance of the main NER implementations on a variety of datasets with differing characteristics (as regards their domain, their size, and their number of classes). We thus provide a deep comparison of algorithms that are never considered together. Our experiments shed some light on how the characteristics of datasets affect the behavior of the methods that we compare.
Despite significant advancements in multi-label text classification, the ability of existing models to generalize to novel and seldom-encountered complex concepts, which are compositions of elementary ones, remains underexplored. This research addresses this gap. By creating unique data splits across three benchmarks, we assess the compositional generalization ability of existing multi-label text classification models. Our results show that these models often fail to generalize to compositional concepts encountered infrequently during training, leading to inferior performance on tests with these new combinations. To address this, we introduce a data augmentation method that leverages two innovative text generation models designed to enhance the classification models' capacity for compositional generalization. Our experiments show that this data augmentation approach significantly improves the compositional generalization capabilities of classification models on our benchmarks, with both generation models surpassing other text generation baselines.
New York City (NYC) topped the global chart for the worst air pollution in June 2023, owing to the wildfire smoke drifting in from Canada. This unprecedented situation caused significant travel disruptions and shifts in traditional activity patterns of NYC residents. This study utilized large-scale social media data to study different crisis activity concerns (i.e., evacuation, staying indoors, shopping, and recreational activities among others) in the emergence of the 2023 Canadian wildfire smoke in NYC. In this regard, one week (June 02 through June 09, 2023) geotagged Twitter data from NYC were retrieved and used in the analysis. The tweets were processed using advanced text classification techniques and later integrated with national databases such as Social Security Administration data, Census, and American Community Survey. Finally, a model has been developed to make community inferences of different activity concerns in a major wildfire. The findings suggest, during wildfires, females are less likely to engage in discussions about evacuation, trips for medical, social, or recreational purposes, and commuting for work, likely influenced by workplaces maintaining operations despite poor air quality. There were also racial disparities in these discussions, with Asians being more likely than Hispanics to discuss evacuation and work commute, and African Americans being less likely to discuss social and recreational activities. Additionally, individuals from low-income neighborhoods and non-higher education students expressed fewer concerns about evacuation. This study provides valuable insights for policymakers, emergency planners, and public health officials, aiding them in formulating targeted communication strategies and equitable emergency response plans.
This study proposes a novel method to understand the factors affecting individuals' perception of transport accessibility, socioeconomic disparity, and public infrastructure. As opposed to the time consuming and expensive survey-based approach, this method can generate organic large-scale responses from social media and develop statistical models to understand individuals' perceptions of various transportation issues. This study retrieved and analyzed 36,098 tweets from New York City from March 19, 2020, to May 15, 2022. A state-of-the-art natural language processing algorithm is used for text mining and classification. A data fusion technique has been adopted to generate a series of socioeconomic traits that are used as explanatory variables in the model. The model results show that females and individuals of Asian origin tend to discuss transportation accessibility more than their counterparts, with those experiencing high neighborhood traffic also being more vocal. However, disadvantaged individuals, including the unemployed and those living in low-income neighborhoods or in areas with high natural hazard risks, tend to communicate less about such issues. As for socioeconomic disparity, individuals of Asian origin and those experiencing various types of air pollution are more likely to discuss these topics on Twitter, often with a negative sentiment. However, unemployed, or disadvantaged individuals, as well as those living in areas with high natural hazard risks or expected losses, are less inclined to tweet about this subject. Lack of internet accessibility could be a reason why many disadvantaged individuals do not tweet about transport accessibility and subsidized internet could be a possible solution.
Nowadays, scene text recognition has attracted more and more attention due to its diverse applications. Most state-of-the-art methods adopt an encoder-decoder framework with the attention mechanism, autoregressively generating text from left to right. Despite the convincing performance, this sequential decoding strategy constrains inference speed. Conversely, non-autoregressive models provide faster, simultaneous predictions but often sacrifice accuracy. Although utilizing an explicit language model can improve performance, it burdens the computational load. Besides, separating linguistic knowledge from vision information may harm the final prediction. In this paper, we propose an alternative solution, using a parallel and iterative decoder that adopts an easy-first decoding strategy. Furthermore, we regard text recognition as an image-based conditional text generation task and utilize the discrete diffusion strategy, ensuring exhaustive exploration of bidirectional contextual information. Extensive experiments demonstrate that the proposed approach achieves superior results on the benchmark datasets, including both Chinese and English text images.
Existing research on audio classification faces challenges in recognizing attributes of passive underwater vessel scenarios and lacks well-annotated datasets due to data privacy concerns. In this study, we introduce CLAPP (Contrastive Language-Audio Pre-training in Passive Underwater Vessel Classification), a novel model. Our aim is to train a neural network using a wide range of vessel audio and vessel state text pairs obtained from an oceanship dataset. CLAPP is capable of directly learning from raw vessel audio data and, when available, from carefully curated labels, enabling improved recognition of vessel attributes in passive underwater vessel scenarios. Model's zero-shot capability allows predicting the most relevant vessel state description for a given vessel audio, without directly optimizing for the task. Our approach aims to solve 2 challenges: vessel audio-text classification and passive underwater vessel audio attribute recognition. The proposed method achieves new state-of-the-art results on both Deepship and Shipsear public datasets, with a notable margin of about 7%-13% for accuracy compared to prior methods on zero-shot task.
The exponential growth of large language models (LLMs) has opened up numerous possibilities for multimodal AGI systems. However, the progress in vision and vision-language foundation models, which are also critical elements of multi-modal AGI, has not kept pace with LLMs. In this work, we design a large-scale vision-language foundation model (InternVL), which scales up the vision foundation model to 6 billion parameters and progressively aligns it with the LLM, using web-scale image-text data from various sources. This model can be broadly applied to and achieve state-of-the-art performance on 32 generic visual-linguistic benchmarks including visual perception tasks such as image-level or pixel-level recognition, vision-language tasks such as zero-shot image/video classification, zero-shot image/video-text retrieval, and link with LLMs to create multi-modal dialogue systems. It has powerful visual capabilities and can be a good alternative to the ViT-22B. We hope that our research could contribute to the development of multi-modal large models. Code and models are available at https://github.com/OpenGVLab/InternVL.