Many high-level skills that are required for computer vision tasks, such as parsing questions, comparing and contrasting semantics, and writing descriptions, are also required in other domains such as natural language processing. In this paper, we ask whether this makes it possible to learn those skills from text data and then use them to complete vision tasks without ever training on visual training data. Key to our approach is exploiting the joint embedding space of contrastively trained vision and language encoders. In practice, there can be systematic differences between embedding spaces for different modalities in contrastive models, and we analyze how these differences affect our approach and study a variety of strategies to mitigate this concern. We produce models using only text training data on three tasks: image captioning, visual entailment and visual question answering, and evaluate them on standard benchmarks using images. We find that this kind of transfer is possible and results in only a small drop in performance relative to models trained on images. We also showcase a variety of stylistic image captioning models that were trained using no image data and no human-curated language data, but instead text data from books, the web, or language models.
In this research, we aim to explore the potential of natural language models (NLMs) such as GPT-3 and GPT-2 to generate effective phishing emails. Phishing emails are fraudulent messages that aim to trick individuals into revealing sensitive information or taking actions that benefit the attackers. We propose a framework for evaluating the performance of NLMs in generating these types of emails based on various criteria, including the quality of the generated text, the ability to bypass spam filters, and the success rate of tricking individuals. Our evaluations show that NLMs are capable of generating phishing emails that are difficult to detect and that have a high success rate in tricking individuals, but their effectiveness varies based on the specific NLM and training data used. Our research indicates that NLMs could have a significant impact on the prevalence of phishing attacks and emphasizes the need for further study on the ethical and security implications of using NLMs for malicious purposes.
This work addresses an alternative approach for query expansion (QE) using a generative adversarial network (GAN) to enhance the effectiveness of information search in e-commerce. We propose a modified QE conditional GAN (mQE-CGAN) framework, which resolves keywords by expanding the query with a synthetically generated query that proposes semantic information from text input. We train a sequence-to-sequence transformer model as the generator to produce keywords and use a recurrent neural network model as the discriminator to classify an adversarial output with the generator. With the modified CGAN framework, various forms of semantic insights gathered from the query document corpus are introduced to the generation process. We leverage these insights as conditions for the generator model and discuss their effectiveness for the query expansion task. Our experiments demonstrate that the utilization of condition structures within the mQE-CGAN framework can increase the semantic similarity between generated sequences and reference documents up to nearly 10% compared to baseline models
Recently, the semantics of scene text has been proven to be essential in fine-grained image classification. However, the existing methods mainly exploit the literal meaning of scene text for fine-grained recognition, which might be irrelevant when it is not significantly related to objects/scenes. We propose an end-to-end trainable network that mines implicit contextual knowledge behind scene text image and enhance the semantics and correlation to fine-tune the image representation. Unlike the existing methods, our model integrates three modalities: visual feature extraction, text semantics extraction, and correlating background knowledge to fine-grained image classification. Specifically, we employ KnowBert to retrieve relevant knowledge for semantic representation and combine it with image features for fine-grained classification. Experiments on two benchmark datasets, Con-Text, and Drink Bottle, show that our method outperforms the state-of-the-art by 3.72\% mAP and 5.39\% mAP, respectively. To further validate the effectiveness of the proposed method, we create a new dataset on crowd activity recognition for the evaluation. The source code and new dataset of this work are available at https://github.com/lanfeng4659/KnowledgeMiningWithSceneText.
We present an in-depth evaluation of four commercially available Speech-to-Text (STT) systems for Swiss German. The systems are anonymized and referred to as system a-d in this report. We compare the four systems to our STT model, referred to as FHNW from hereon after, and provide details on how we trained our model. To evaluate the models, we use two STT datasets from different domains. The Swiss Parliament Corpus (SPC) test set and a private dataset in the news domain with an even distribution across seven dialect regions. We provide a detailed error analysis to detect the three systems' strengths and weaknesses. This analysis is limited by the characteristics of the two test sets. Our model scored the highest bilingual evaluation understudy (BLEU) on both datasets. On the SPC test set, we obtain a BLEU score of 0.607, whereas the best commercial system reaches a BLEU score of 0.509. On our private test set, we obtain a BLEU score of 0.722 and the best commercial system a BLEU score of 0.568.
Training machines to understand natural language and interact with humans is one of the major goals of artificial intelligence. Recent years have witnessed an evolution from matching networks to pre-trained language models (PrLMs). In contrast to the plain-text modeling as the focus of the PrLMs, dialogue texts involve multiple speakers and reflect special characteristics such as topic transitions and structure dependencies between distant utterances. However, the related PrLM models commonly represent dialogues sequentially by processing the pairwise dialogue history as a whole. Thus the hierarchical information on either utterance interrelation or speaker roles coupled in such representations is not well addressed. In this work, we propose compositional learning for holistic interaction across the utterances beyond the sequential contextualization from PrLMs, in order to capture the utterance-aware and speaker-aware representations entailed in a dialogue history. We decouple the contextualized word representations by masking mechanisms in Transformer-based PrLM, making each word only focus on the words in current utterance, other utterances, and two speaker roles (i.e., utterances of sender and utterances of the receiver), respectively. In addition, we employ domain-adaptive training strategies to help the model adapt to the dialogue domains. Experimental results show that our method substantially boosts the strong PrLM baselines in four public benchmark datasets, achieving new state-of-the-art performance over previous methods.
Multi-modal named entity recognition (NER) and relation extraction (RE) aim to leverage relevant image information to improve the performance of NER and RE. Most existing efforts largely focused on directly extracting potentially useful information from images (such as pixel-level features, identified objects, and associated captions). However, such extraction processes may not be knowledge aware, resulting in information that may not be highly relevant. In this paper, we propose a novel Multi-modal Retrieval based framework (MoRe). MoRe contains a text retrieval module and an image-based retrieval module, which retrieve related knowledge of the input text and image in the knowledge corpus respectively. Next, the retrieval results are sent to the textual and visual models respectively for predictions. Finally, a Mixture of Experts (MoE) module combines the predictions from the two models to make the final decision. Our experiments show that both our textual model and visual model can achieve state-of-the-art performance on four multi-modal NER datasets and one multi-modal RE dataset. With MoE, the model performance can be further improved and our analysis demonstrates the benefits of integrating both textual and visual cues for such tasks.
Language models (LMs) are trained on collections of documents, written by individual human agents to achieve specific goals in an outside world. During training, LMs have access only to text of these documents, with no direct evidence of the internal states of the agents that produced them -- a fact often used to argue that LMs are incapable of modeling goal-directed aspects of human language production and comprehension. Can LMs trained on text learn anything at all about the relationship between language and use? I argue that LMs are models of intentional communication in a specific, narrow sense. When performing next word prediction given a textual context, an LM can infer and represent properties of an agent likely to have produced that context. These representations can in turn influence subsequent LM generation in the same way that agents' communicative intentions influence their language. I survey findings from the recent literature showing that -- even in today's non-robust and error-prone models -- LMs infer and use representations of fine-grained communicative intentions and more abstract beliefs and goals. Despite the limited nature of their training data, they can thus serve as building blocks for systems that communicate and act intentionally.
As multimodal learning finds applications in a wide variety of high-stakes societal tasks, investigating their robustness becomes important. Existing work has focused on understanding the robustness of vision-and-language models to imperceptible variations on benchmark tasks. In this work, we investigate the robustness of multimodal classifiers to cross-modal dilutions - a plausible variation. We develop a model that, given a multimodal (image + text) input, generates additional dilution text that (a) maintains relevance and topical coherence with the image and existing text, and (b) when added to the original text, leads to misclassification of the multimodal input. Via experiments on Crisis Humanitarianism and Sentiment Detection tasks, we find that the performance of task-specific fusion-based multimodal classifiers drops by 23.3% and 22.5%, respectively, in the presence of dilutions generated by our model. Metric-based comparisons with several baselines and human evaluations indicate that our dilutions show higher relevance and topical coherence, while simultaneously being more effective at demonstrating the brittleness of the multimodal classifiers. Our work aims to highlight and encourage further research on the robustness of deep multimodal models to realistic variations, especially in human-facing societal applications. The code and other resources are available at https://claws-lab.github.io/multimodal-robustness/.
With the development of deep learning and Transformer-based pre-trained models like BERT, the accuracy of many NLP tasks has been dramatically improved. However, the large number of parameters and computations also pose challenges for their deployment. For instance, using BERT can improve the predictions in the financial sentiment analysis (FSA) task but slow it down, where speed and accuracy are equally important in terms of profits. To address these issues, we first propose an efficient and lightweight BERT (ELBERT) along with a novel confidence-window-based (CWB) early exit mechanism. Based on ELBERT, an innovative method to accelerate text processing on the GPU platform is developed, solving the difficult problem of making the early exit mechanism work more effectively with a large input batch size. Afterward, a fast and high-accuracy FSA system is built. Experimental results show that the proposed CWB early exit mechanism achieves significantly higher accuracy than existing early exit methods on BERT under the same computation cost. By using this acceleration method, our FSA system can boost the processing speed by nearly 40 times to over 1000 texts per second with sufficient accuracy, which is nearly twice as fast as FastBERT, thus providing a more powerful text processing capability for modern trading systems.