Emotion recognition in conversation (ERC) is an important topic for developing empathetic machines a variety of areas including social opinion mining, health-care and so on. In this paper, we propose method to model ERC task as sequence tagging where Conditional Random Field (CRF) layer leveraged learn the emotional consistency conversation. We employ LSTM-based encoders that capture self inter-speaker dependency interlocutors generate contextualized utterance representations which are fed...

We treat streamline selection and viewpoint as symmetric problems which are formulated into a unified information-theoretic framework. This is achieved by building two interrelated information channels between pool of candidate streamlines set sample viewpoints. define the to select best in similar manner, Furthermore, we propose solutions clustering partitioning based on representativeness viewpoints, respectively. Finally, camera path that passes through all selected viewpoints for...

Span-based joint extraction models have shown their efficiency on entity recognition and relation extraction. These regard text spans as candidate entities span tuples tuples. Span semantic representations are shared in both extraction, while existing cannot well capture semantics of these relations. To address problems, we introduce a span-based framework with attention-based representations. Specially, attentions utilized to calculate representations, including span-specific contextual...

Multilevel thresholding segmentation of color images is an important technology in various applications which has received more attention recent years. The process determining the optimal threshold values case traditional methods time-consuming. In order to mitigate above problem, meta-heuristic algorithms have been employed this field for searching optima during past few paper, effective technique Electromagnetic Field Optimization (EFO) algorithm based on a fuzzy entropy criterion...

We aim at summarizing answers in community question-answering (CQA). While most previous work focuses on factoid question-answering, we focus the non-factoid question-answering. Unlike CQA, usually requires passages as answers. The shortness, sparsity and diversity of form interesting challenges for summarization. To tackle these challenges, propose a sparse coding-based summarization strategy that includes three core ingredients: short document expansion, sentence vectorization,...

Code search aims to for relevant code snippets through queries, which has become an essential requirement assist programmers in software development. With the availability of large and rapidly growing source repositories covering various languages, multilingual can leverage more training data learn complementary information across languages. Contrastive learning naturally understand similarity between functionally equivalent different languages by narrowing distance objects with same...

Wireless sensor networks (WSNs) are widely implemented in military, intelligent medical, transportation, space exploration, and other fields. However, the authentication communication of WSNs carried out harsh external environment via a public channel, which is more vulnerable to various attacks than traditional networks. Authentication key technology security measures, but common scheme not suitable for due limitations memory, computing, energy consumption. Therefore, designing secure...

With an increasingly amount of information in web forums, quick comprehension threads forums has become a challenging research problem. To handle this issue, paper investigates the task Web Forum Thread Summarization (WFTS), aiming to give brief statement each thread that involving multiple dynamic topics. When applied WFTS, traditional summarization methods are cramped by topic dependencies, drifting and text sparseness. Consequently, we explore unsupervised propagation model paper, Post...

Streamline visualization can be formulated as the problem of streamline placement or selection. In this paper, we present an importance-driven approach to view-dependent selection that guarantees coherent update when view changes gradually. Given a large number randomly uniformly seeded and traced streamlines sample viewpoints, our evaluates, for each streamline, importance by considering amount information shared 3D its 2D projection well how stereoscopic streamline's shape is reflected...

Topic-specific political stance inference in social networks (SNs) aims at inferring target users' attitudes toward different topics. Traditional methods mainly used a language model to classify sentiments from the postings of SN users. However, people's stances are not always equal their sentiments. Some others tried build separate models In many cases, though users talked about topics, information given was limited; or they only expressed some other issues except When is incomplete, that...

The main computational steps in alge- braic soft-decoding of Reed-Solomon codes are bivari- ate interpolation and factorization. We discuss a new technique, based upon re-encoding coordinate transformation, that significantly reduces the complexity procedure transfor- mation convert original problem into another reduced problem, which is orders magnitude smaller than one.

Previous chapter Next Full AccessProceedings Proceedings of the 2012 SIAM International Conference on Data Mining (SDM)Transfer Topic Modeling with Ease and ScalabilityJeon-Hyung Kang, Jun Ma, Yan LiuJeon-Hyung Liupp.564 - 575Chapter DOI:https://doi.org/10.1137/1.9781611972825.49PDFBibTexSections ToolsAdd to favoritesExport CitationTrack CitationsEmail SectionsAboutAbstract The increasing volume short texts generated social media sites, such as Twitter or Facebook, creates a great demand for...

Each cloud service has numerous owners and tenants, so it is necessary to construct a privacy preserving identity management access control mechanism for computing.On one hand, providers (CSP) depend on tenant's information enforce appropriate that resources are only accessed by the authorized tenants who willing pay.On other wish protect their personalized patterns, accessing newfangled services on-demand ways within scope of permissions.There many authentication schemes address these...

Unstructured Persona-oriented Dialogue Systems (UPDS) has been demonstrated effective in generating persona consistent responses by utilizing predefined natural language user descriptions (e.g., "I am a vegan"). However, the are usually short and limited to only few descriptive words, which makes it hard correlate them with dialogues. As result, existing methods either fail use description or improperly when responses. To address this, we propose neural topical expansion framework, namely...