Recent years have witnessed the boom of online sharing media contents, which raise significant challenges in effective management and retrieval. Though a large amount efforts been made, precise retrieval on video shots with certain topics has largely ignored. At same time, due to popularity novel time-sync comments, or so-called "bullet-screen comments", semantics could be now combined timestamps support further research temporal labeling. In this paper, we propose understanding framework...

Aspect sentiment classification (ASC) is a fundamental task in analysis. It aims at classifying the expressed on some target aspects/features of entities (e.g., products and services). Although great deal research has been done, this remains to be very challenging. Recently, memory networks, type neural model, have used for achieved state-of-the-art results. However, such models usually require large amount well-annotated training data producing reasonably good Unfortunately, ASC task,...

Sentence semantic matching requires an agent to determine the relation between two sentences, which is widely used in various natural language tasks such as Natural Language Inference (NLI) and Paraphrase Identification (PI). Among all methods, attention mechanism plays important role capturing relations properly aligning elements of sentences. Previous methods utilized select parts sentences at one time. However, sentence during are dynamically changing with degree understanding. Selecting...

Sentence semantic matching is one of the fundamental tasks in natural language processing, which requires an agent to determine relation among input sentences. Recently, deep neural networks have achieved impressive performance this area, especially BERT. Despite effectiveness these models, most them treat output labels as meaningless one-hot vectors, underestimating information and guidance relations that reveal, for with a small number labels. To address problem, we propose Relation...

Text classification is one of the fundamental tasks in natural language processing, which requires an agent to determine most appropriate category for input sentences. Recently, deep neural networks have achieved impressive performance this area, especially pretrained models (PLMs). Usually, these methods concentrate on sentences and corresponding semantic embedding generation. However, another essential component: labels, existing works either treat them as meaningless one-hot vectors or...

Recognizing lexical entailment (RLE) always plays an important role in inference of natural language, i.e., identifying whether one word entails another, for example, fox animal. In the literature, automatically recognizing pairs deeply relies on words' contextual representations. However, as a "prototype" vector, single representation cannot reveal multifaceted aspects words due to their homonymy and polysemy. this paper, we propose supervised Context-Enriched Neural Network (CENN) method...

With the booming of content "re-creation" in social media platforms, character-orientedvideo summary has become a crucial form user-generated video content. However, artificial extraction could be time-consuming with high missing rate, while traditional techniques on person search may incur heavy burden computing resources. At same time, videos are usually accompanied rich textual information, e.g., subtitles or bullet-screen comments which provide multi-view description videos. Thus, there...

Visual Emotion Recognition has attracted more and research attention in recent years. Existing approaches mainly depend on facial expression or analyze the whole image between positive negative. Actually, people can recognize multiple emotions from one based global 10-cal information. In this paper, we propose a Context-Aware Generation-Based Net (CAGBN), novel architecture that makes full use of local information by considering both details target person. Inspired psychological studies when...

Natural Language Inference (NLI) task requires an agent to determine the semantic relation between a premise sentence (p) and hypothesis (h), which demands sufficient understanding about sentences from lexical knowledge global semantic. Due issues such as polysemy, ambiguity, well fuzziness of sentences, fully is still challenging. To this end, we propose Image-Enhanced Multi-Level Sentence Representation Net (IEMLRN), novel architecture that able utilize image enhance at different scales....

Recent years have witnessed a successful rise of the time synchronized <i>gossiping comment</i>, or so-called danmu combined with online videos. This new business mode has enriched communication among users by sending users&#x2019; feelings through danmus and sharing these on Can be helpful for better user behavior modeling video analyzing? To this question, in article, preliminary attempts are made analysis videos introducing <i>Danmu</i> dataset which is collected from real-world...

Natural language inference (NLI) task requires an agent to determine the semantic relation between a premise sentence ( p) and hypothesis h), which demands sufficient understanding about sentences semantic. Due issues, such as polysemy, ambiguity, well fuzziness of sentences, intense is very challenging. To this end, in article, we introduce corresponding image reference information for enhancing representing. Specifically, first propose image-enhanced multilevel representation net (IEMLRN),...

Sentiment lexicons are instrumental for sentiment analysis.One can use a set of words provided in lexicon and lexicon-based classifier to perform major issue with this approach is that many (from the lexicon) domain dependent.That is, they may be positive some domains but negative others.We refer problem as polarity-changes from lexicon.Detecting such correcting their an application very important.In paper, we propose graph-based technique tackle problem.Experimental results show its...

Sentence semantic matching requires an agent to determine the relation between two sentences, which is widely used in various natural language tasks, such as inference (NLI) and paraphrase identification (PI). Much recent progress has been made this area, especially attention-based methods pretrained model-based methods. However, most of these focus on all important parts sentences a static way only emphasize how words are query, inhibiting ability attention mechanism. In order overcome...

Technological change and innovation are vitally important, especially for high-tech companies. However, factors influencing their future research development (R&D) trends both complicated various, leading it a quite difficult task to make technology tracing To this end, in paper, we develop novel data-driven solution, i.e., Deep Technology Forecasting (DTF) framework, automatically find the most possible directions customized each company. Specially, DTF consists of three components:...

Although deep learning has demonstrated its outstanding performance on image classification, most well-known networks make efforts to optimize both their structures and node weights for recognizing fewer (e.g., no more than 1000) object classes. Therefore, it is attractive extend or mixture such support large-scale classification. According our best knowledge, how adaptively effectively fuse multiple CNNs classification still under-explored. On this basis, a algorithm developed in paper....

Text Classification is one of the fundamental tasks in natural language processing, which requires an agent to determine most appropriate category for input sentences. Recently, deep neural networks have achieved impressive performance this area, especially Pre-trained Language Models (PLMs). Usually, these methods concentrate on sentences and corresponding semantic embedding generation. However, another essential component: labels, existing works either treat them as meaningless one-hot...