In this paper, we propose a novel architecture for multi-modal speech and text input. We combine pretrained speech and text encoders using multi-headed cross-modal attention and jointly fine-tune on the target problem. The resultant architecture can be used for continuous token-level classification or utterance-level prediction acting on simultaneous text and speech. The resultant encoder efficiently captures both acoustic-prosodic and lexical information. We compare the benefits of multi-headed attention-based fusion for multi-modal utterance-level classification against a simple concatenation of pre-pooled, modality-specific representations. Our model architecture is compact, resource efficient, and can be trained on a single consumer GPU card.
Stock market movements are influenced by public and private information shared through news articles, company reports, and social media discussions. Analyzing these vast sources of data can give market participants an edge to make profit. However, the majority of the studies in the literature are based on traditional approaches that come short in analyzing unstructured, vast textual data. In this study, we provide a review on the immense amount of existing literature of text-based stock market analysis. We present input data types and cover main textual data sources and variations. Feature representation techniques are then presented. Then, we cover the analysis techniques and create a taxonomy of the main stock market forecast models. Importantly, we discuss representative work in each category of the taxonomy, analyzing their respective contributions. Finally, this paper shows the findings on unaddressed open problems and gives suggestions for future work. The aim of this study is to survey the main stock market analysis models, text representation techniques for financial market prediction, shortcomings of existing techniques, and propose promising directions for future research.
We propose to tackle conditional text generation tasks, especially those which require generating formulaic text, by splicing together segments of text from retrieved "neighbor" source-target pairs. Unlike recent work that conditions on retrieved neighbors in an encoder-decoder setting but generates text token-by-token, left-to-right, we learn a policy that directly manipulates segments of neighbor text (i.e., by inserting or replacing them) to form an output. Standard techniques for training such a policy require an oracle derivation for each generation, and we prove that finding the shortest such derivation can be reduced to parsing under a particular weighted context-free grammar. We find that policies learned in this way allow for interpretable table-to-text or headline generation that is competitive with neighbor-based token-level policies on automatic metrics, though on all but one dataset neighbor-based policies underperform a strong neighborless baseline. In all cases, however, generating by splicing is faster.
With the tremendously increasing number of videos, there is a great demand for techniques that help people quickly navigate to the video segments they are interested in. However, current works on video understanding mainly focus on video content summarization, while little effort has been made to explore the structure of a video. Inspired by textual outline generation, we introduce a novel video understanding task, namely video outline generation (VOG). This task is defined to contain two sub-tasks: (1) first segmenting the video according to the content structure and then (2) generating a heading for each segment. To learn and evaluate VOG, we annotate a 10k+ dataset, called DuVOG. Specifically, we use OCR tools to recognize subtitles of videos. Then annotators are asked to divide subtitles into chapters and title each chapter. In videos, highlighted text tends to be the headline since it is more likely to attract attention. Therefore we propose a Visual Subtitle feature Enhanced video outline generation model (VSENet) which takes as input the textual subtitles together with their visual font sizes and positions. We consider the VOG task as a sequence tagging problem that extracts spans where the headings are located and then rewrites them to form the final outlines. Furthermore, based on the similarity between video outlines and textual outlines, we use a large number of articles with chapter headings to pretrain our model. Experiments on DuVOG show that our model largely outperforms other baseline methods, achieving 77.1 of F1-score for the video segmentation level and 85.0 of ROUGE-L_F0.5 for the headline generation level.
Text lines are important parts of handwritten document images and easier to analyze by further applications. Despite recent progress in text line detection, text line extraction from a handwritten document remains an unsolved task. This paper proposes to use a fully convolutional network for text line detection and energy minimization for text line extraction. Detected text lines are represented by blob lines that strike through the text lines. These blob lines assist an energy function for text line extraction. The detection stage can locate arbitrarily oriented text lines. Furthermore, the extraction stage is capable of finding out the pixels of text lines with various heights and interline proximity independent of their orientations. Besides, it can finely split the touching and overlapping text lines without an orientation assumption. We evaluate the proposed method on VML-AHTE, VML-MOC, and Diva-HisDB datasets. The VML-AHTE dataset contains overlapping, touching and close text lines with rich diacritics. The VML-MOC dataset is very challenging by its multiply oriented and skewed text lines. The Diva-HisDB dataset exhibits distinct text line heights and touching text lines. The results demonstrate the effectiveness of the method despite various types of challenges, yet using the same parameters in all the experiments.
Large-scale auto-regressive language models pretrained on massive text have demonstrated their impressive ability to perform new natural language tasks with only a few text examples, without the need for fine-tuning. Recent studies further show that such a few-shot learning ability can be extended to the text-image setting by training an encoder to encode the images into embeddings functioning like the text embeddings of the language model. Interested in exploring the possibility of transferring the few-shot learning ability to the audio-text setting, we propose a novel speech understanding framework, WavPrompt, where we finetune a wav2vec model to generate a sequence of audio embeddings understood by the language model. We show that WavPrompt is a few-shot learner that can perform speech understanding tasks better than a naive text baseline. We conduct detailed ablation studies on different components and hyperparameters to empirically identify the best model configuration. In addition, we conduct a non-speech understanding experiment to show WavPrompt can extract more information than just the transcriptions.
The novel coronavirus disease (COVID-19) began in Wuhan, China, in late 2019 and to date has infected over 148M people worldwide, resulting in 3.12M deaths. On March 10, 2020, the World Health Organisation (WHO) declared it as a global pandemic. Many academicians and researchers started to publish papers describing the latest discoveries on covid-19. The large influx of publications made it hard for other researchers to go through a large amount of data and find the appropriate one that helps their research. So, the proposed model attempts to extract relavent titles from the large corpus of research publications which makes the job easy for the researchers. Allen Institute for AI released the CORD-19 dataset, which consists of 2,00,000 journal articles related to coronavirus-related research publications from PubMed's PMC, WHO (World Health Organization), bioRxiv, and medRxiv pre-prints. Along with this document corpus, they have also provided a topics dataset named topics-rnd3 consisting of a list of topics. Each topic has three types of representations like query, question, and narrative. These Datasets are made open for research, and also they released a TREC-COVID competition on Kaggle. Using these topics like queries, our goal is to find out the relevant documents in the CORD-19 dataset. In this research, relevant documents should be recognized for the posed topics in topics-rnd3 data set. The proposed model uses Natural Language Processing(NLP) techniques like Bag-of-Words, Average Word-2-Vec, Average BERT Base model and Tf-Idf weighted Word2Vec model to fabricate vectors for query, question, narrative, and combinations of them. Similarly, fabricate vectors for titles in the CORD-19 dataset. After fabricating vectors, cosine similarity is used for finding similarities between every two vectors. Cosine similarity helps us to find relevant documents for the given topic.
Text detection, the key technology for understanding scene text, has become an attractive research topic. For detecting various scene texts, researchers propose plenty of detectors with different advantages: detection-based models enjoy fast detection speed, and segmentation-based algorithms are not limited by text shapes. However, for most intelligent systems, the detector needs to detect arbitrary-shaped texts with high speed and accuracy simultaneously. Thus, in this study, we design an efficient pipeline named as MT, which can detect adhesive arbitrary-shaped texts with only a single binary mask in the inference stage. This paper presents the contributions on three aspects: (1) a light-weight detection framework is designed to speed up the inference process while keeping high detection accuracy; (2) a multi-perspective feature module is proposed to learn more discriminative representations to segment the mask accurately; (3) a multi-factor constraints IoU minimization loss is introduced for training the proposed model. The effectiveness of MT is evaluated on four real-world scene text datasets, and it surpasses all the state-of-the-art competitors to a large extent.
Intent detection of spoken queries is a challenging task due to their noisy structure and short length. To provide additional information regarding the query and enhance the performance of intent detection, we propose a method for semantic expansion of spoken queries, called ConQX, which utilizes the text generation ability of an auto-regressive language model, GPT-2. To avoid off-topic text generation, we condition the input query to a structured context with prompt mining. We then apply zero-shot, one-shot, and few-shot learning. We lastly use the expanded queries to fine-tune BERT and RoBERTa for intent detection. The experimental results show that the performance of intent detection can be improved by our semantic expansion method.
An emerging theme in artificial intelligence research is the creation of models to simulate the decisions and behavior of specific people, in domains including game-playing, text generation, and artistic expression. These models go beyond earlier approaches in the way they are tailored to individuals, and the way they are designed for interaction rather than simply the reproduction of fixed, pre-computed behaviors. We refer to these as mimetic models, and in this paper we develop a framework for characterizing the ethical and social issues raised by their growing availability. Our framework includes a number of distinct scenarios for the use of such models, and considers the impacts on a range of different participants, including the target being modeled, the operator who deploys the model, and the entities that interact with it.