Meibomian Gland Dysfunction (MGD) is a key contributor to clinical Dry Eye Disease (DED), and accurate meibomian gland segmentation is essential for its intelligent diagnosis. However, the high cost of annotation remains a major barrier to improving segmentation quality. Active Learning (AL), which selects the most informative samples for annotation, offers an efficient solution. This study explores the application of AL to infrared meibography image segmentation and introduces a novel progressive hybrid sampling framework. Initial AL stages often rely on random sampling for initial labeled sets, risking unstable performance from uninformative data and cold starts. Additionally, meibography images always have specular reflections whose late introduction in training degrades future segmentation accuracy. To address these challenges, we incorporate prior knowledge of specular reflections into the initial labeled set construction, enabling the model to handle such artifacts from the beginning. We then implement a two-stage dynamic sampling strategy: entropy-based uncertainty sampling is used in the early iterations to maximize annotated data informativeness and rapidly boost model performance. However, as AL progresses, only focusing on uncertainty may overlook global data distribution and lead to redundant annotations. Hence, an adaptive threshold is introduced to monitor sample redundancy. When redundancy exceeds this threshold, a diversity sampling module is activated to improve model generalization. Experimental results on the public MGD-1K dataset show that our method achieves superior segmentation performance under the same annotation budget. Unlike mainstream methods that often fail to outperform random baselines, our approach consistently delivers accurate and efficient meibomian gland segmentation.

Low-cost and high-precision infrared meibomian gland segmentation is an important basis for early diagnosis and monitoring of many ocular diseases in ophthalmic clinical practice. To address the issue of limited labeled data, we propose a novel semi-supervised meibomian gland segmentation approach. By leveraging the prior knowledge of the patient each image belongs to, intra-patient registration is taken to generate diverse and lifelike pseudo-labeled data. Contrastive learning strategy with reliable negative sample filtering is also introduced to resolve the insufficient supervision in the feature space. Experimental results on the private dataset demonstrate the success of the proposed approach, exhibiting superiority over the state-of-the-art methods. The Code can be found at SS-MGS.

We introduce a framework of Monte Carlo Tree Search with Double-q Dueling network (MCTS-DDU) for task-completion dialogue policy learning. Different from the previous deep model-based reinforcement learning methods, which uses background planning and may suffer from low-quality simulated experiences, MCTS-DDU performs decision-time planning based on dialogue state search trees built by Monte Carlo simulations and is robust to the simulation errors. Such idea arises naturally in human behaviors, e.g. predicting others' responses and then deciding our own actions. In the simulated movie-ticket booking task, our method outperforms the background planning approaches significantly. We demonstrate the effectiveness of MCTS and the dueling network in detailed ablation studies, and also compare the performance upper bounds of these two planning methods.

Extracting events accurately from vast news corpora and organize events logically is critical for news apps and search engines, which aim to organize news information collected from the Internet and present it to users in the most sensible forms. Intuitively speaking, an event is a group of news documents that report the same news incident possibly in different ways. In this article, we describe our experience of implementing a news content organization system at Tencent to discover events from vast streams of breaking news and to evolve news story structures in an online fashion. Our real-world system faces unique challenges in contrast to previous studies on topic detection and tracking (TDT) and event timeline or graph generation, in that we (1) need to accurately and quickly extract distinguishable events from massive streams of long text documents, and (2) must develop the structures of event stories in an online manner, in order to guarantee a consistent user viewing experience. In solving these challenges, we propose Story Forest , a set of online schemes that automatically clusters streaming documents into events, while connecting related events in growing trees to tell evolving stories. A core novelty of our Story Forest system is EventX , a semi-supervised scheme to extract events from massive Internet news corpora. EventX relies on a two-layered, graph-based clustering procedure to group documents into fine-grained events. We conducted extensive evaluations based on (1) 60 GB of real-world Chinese news data, (2) a large Chinese Internet news dataset that contains 11,748 news articles with truth event labels, and (3) the 20 News Groups English dataset, through detailed pilot user experience studies. The results demonstrate the superior capabilities of Story Forest to accurately identify events and organize news text into a logical structure that is appealing to human readers.

Identifying the relationship between two articles, e.g., whether two articles published from different sources describe the same breaking news, is critical to many document understanding tasks.Existing approaches for modeling and matching sentence pairs do not perform well in matching longer documents, which embody more complex interactions between the enclosed entities than a sentence does.To model article pairs, we propose the Concept Interaction Graph to represent an article as a graph of concepts.We then match a pair of articles by comparing the sentences that enclose the same concept vertex through a series of encoding techniques, and aggregate the matching signals through a graph convolutional network.To facilitate the evaluation of long article matching, we have created two datasets, each consisting of about 30K pairs of breaking news articles covering diverse topics in the open domain.Extensive evaluations of the proposed methods on the two datasets demonstrate significant improvements over a wide range of state-of-the-art methods for natural language matching.

Events are happening in real-world and real-time, which can be planned and organized occasions involving multiple people and objects. Social media platforms publish a lot of text messages containing public events with comprehensive topics. However, mining social events is challenging due to the heterogeneous event elements in texts and explicit and implicit social network structures. In this paper, we design an event meta-schema to characterize the semantic relatedness of social events and build an event-based heterogeneous information network (HIN) integrating information from external knowledge base, and propose a novel Pairwise Popularity Graph Convolutional Network (PP-GCN) based fine-grained social event categorization model. We propose a Knowledgeable meta-paths Instances based social Event Similarity (KIES) between events and build a weighted adjacent matrix as input to the PP-GCN model. Comprehensive experiments on real data collections are conducted to compare various social event detection and clustering tasks. Experimental results demonstrate that our proposed framework outperforms other alternative social event categorization techniques.

Related search query recommendation is a standard feature in many modern search engines. Interesting and relevant queries often increase the active time of users and improve the overall search experience. However, conventional approaches based on tag extraction, keywords matching or click graph link analysis suffer from the common problem of limited coverage and generalizability, which means the system could only make suggestions for a small portion of well-formed search queries.

Concepts embody the knowledge of the world and facilitate the cognitive processes of human beings. Mining concepts from web documents and constructing the corresponding taxonomy are core research problems in text understanding and support many downstream tasks such as query analysis, knowledge base construction, recommendation, and search. However, we argue that most prior studies extract formal and overly general concepts from Wikipedia or static web pages, which are not representing the user perspective.

We study the problem of search query inference from web documents, where a short, comprehensive natural language query is inferred from a long article. Search query generation or inference is of great value to search engines and recommenders in terms of locating potential target users and ranking content. Despite being closely related to other NLP tasks like abstract generation and keyword extraction, we point out that search query inference is, in fact, a new problem, in that the generated natural language query, which consists of a few words, is expected to be comprehensive enough to lead to the click-through of the corresponding document. Therefore, query generation requires an accurate inference of query words, as well as a deeper level of understanding on document semantic structures. Toward this end, we propose a novel generative model called the Graph-augmented Sequence to Attention (G-S2A) network. Adopting an Encoder-Decoder architecture, G-S2A incorporates a sentence-level Graph Convolutional Network (GCN), a keyword-level GCN, as well as a hierarchical recurrent neural network (RNN) into the encoder to generate structural document representations. An attentional Transformer decoder is then applied to combine different types of encoded features to generate a target query. On a query-document dataset from a real-world search engine, our model outperforms several neural generative models on a wide range of metrics.

Automatic question generation is an important technique that can improve the training of question answering, help chatbots to start or continue a conversation with humans, and provide assessment materials for educational purposes. Existing neural question generation models are not sufficient mainly due to their inability to properly model the process of how each word in the question is selected, i.e., whether repeating the given passage or being generated from a vocabulary. In this paper, we propose our Clue Guided Copy Network for Question Generation (CGC-QG), which is a sequence-to-sequence generative model with copying mechanism, yet employing a variety of novel components and techniques to boost the performance of question generation. In CGC-QG, we design a multi-task labeling strategy to identify whether a question word should be copied from the input passage or be generated instead, guiding the model to learn the accurate boundaries between copying and generation. Furthermore, our input passage encoder takes as input, among a diverse range of other features, the prediction made by a clue word predictor, which helps identify whether each word in the input passage is a potential clue to be copied into the target question. The clue word predictor is designed based on a novel application of Graph Convolutional Networks onto a syntactic dependency tree representation of each passage, thus being able to predict clue words only based on their context in the passage and their relative positions to the answer in the tree. We jointly train the clue prediction as well as question generation with multi-task learning and a number of practical strategies to reduce the complexity. Extensive evaluations show that our model significantly improves the performance of question generation and out-performs all previous state-of-the-art neural question generation models by a substantial margin.

A large number of deep learning models have been proposed for the text matching problem, which is at the core of various typical natural language processing (NLP) tasks. However, existing deep models are mainly designed for the semantic matching between a pair of short texts, such as paraphrase identification and question answering, and do not perform well on the task of relevance matching between short-long text pairs. This is partially due to the fact that the essential characteristics of short-long text matching have not been well considered in these deep models. More specifically, these methods fail to handle extreme length discrepancy between text pieces and neither can they fully characterize the underlying structural information in long text documents. In this paper, we are especially interested in relevance matching between a piece of short text and a long document, which is critical to problems like query-document matching in information retrieval and web searching. To extract the structural information of documents, an undirected graph is constructed, with each vertex representing a keyword and the weight of an edge indicating the degree of interaction between keywords. Based on the keyword graph, we further propose a Multiresolution Graph Attention Network to learn multi-layered representations of vertices through a Graph Convolutional Network (GCN), and then match the short text snippet with the graphical representation of the document with an attention mechanism applied over each layer of the GCN. Experimental results on two datasets demonstrate that our graph approach outperforms other state-of-the-art deep matching models.

Short Text Matching plays an important role in many natural language processing tasks such as information retrieval, question answering, and conversational system. Conventional text matching methods rely on predefined templates and rules, which are not applicable to short text with limited numebr of words and limit their ability to generalize to unobserved data. Many recent efforts have been made to apply deep neural network models to natural language processing tasks, which reduces the cost of feature engineering. In this paper, we present the design of Multi-Channel Information Crossing , a multi-channel convolutional neural network model for text matching, with additional attention mechanisms from sentence and text semantics. MIX compares text snippets at varied granularities to form a series of multi-channel similarity matrices, which are crossed with another set of carefully designed attention matrices to expose the rich structures of sentences to deep neural networks. We implemented MIX and deployed the system on Tencent's Venus distributed computation platform. Thanks to carefully engineered multi-channel information crossing, evaluation results suggest that MIX outperforms a wide range of state-of-the-art deep neural network models by at least 11.1% in terms of the normalized discounted cumulative gain ([email protected]), on the English WikiQA dataset. Moreover, we also performed online A/B tests with real users on the search service of Tencent QQ Browser. Results suggest that MIX raised the number of clicks on the returned results by 5.7%, due to an increased accuracy in query-document matching, which demonstrates the superior performance of MIX in production environments.

Semantic matching of natural language sentences or identifying the relationship between two sentences is a core research problem underlying many natural language tasks. Depending on whether training data is available, prior research has proposed both unsupervised distance-based schemes and supervised deep learning schemes for sentence matching. However, previous approaches either omit or fail to fully utilize the ordered, hierarchical, and flexible structures of language objects, as well as the interactions between them. In this paper, we propose Hierarchical Sentence Factorization---a technique to factorize a sentence into a hierarchical representation, with the components at each different scale reordered into a "predicate-argument" form. The proposed sentence factorization technique leads to the invention of: 1) a new unsupervised distance metric which calculates the semantic distance between a pair of text snippets by solving a penalized optimal transport problem while preserving the logical relationship of words in the reordered sentences, and 2) new multi-scale deep learning models for supervised semantic training, based on factorized sentence hierarchies. We apply our techniques to text-pair similarity estimation and text-pair relationship classification tasks, based on multiple datasets such as STSbenchmark, the Microsoft Research paraphrase identification (MSRP) dataset, the SICK dataset, etc. Extensive experiments show that the proposed hierarchical sentence factorization can be used to significantly improve the performance of existing unsupervised distance-based metrics as well as multiple supervised deep learning models based on the convolutional neural network (CNN) and long short-term memory (LSTM).

We describe our experience of implementing a news content organization system at Tencent that discovers events from vast streams of breaking news and evolves news story structures in an online fashion. Our real-world system has distinct requirements in contrast to previous studies on topic detection and tracking (TDT) and event timeline or graph generation, in that we 1) need to accurately and quickly extract distinguishable events from massive streams of long text documents that cover diverse topics and contain highly redundant information, and 2) must develop the structures of event stories in an online manner, without repeatedly restructuring previously formed stories, in order to guarantee a consistent user viewing experience. In solving these challenges, we propose Story Forest, a set of online schemes that automatically clusters streaming documents into events, while connecting related events in growing trees to tell evolving stories. We conducted extensive evaluation based on 60 GB of real-world Chinese news data, although our ideas are not language-dependent and can easily be extended to other languages, through detailed pilot user experience studies. The results demonstrate the superior capability of Story Forest to accurately identify events and organize news text into a logical structure that is appealing to human readers, compared to multiple existing algorithm frameworks.

We study the video recommendation problem based on a large amount of user viewing logs instead of explicit ratings. As viewing records are implicitly suggest user preferences, existing matrix factorization methods fail to generate discriminative recommendations based on such one-class positive samples. We propose a scalable approach called separating-plane matrix factorization (SPMF) to make effective recommendations based on positive implicit feedback, with a learning complexity that is comparable to traditional matrix factorization. With extensive offline evaluation in Tencent Data Warehouse (TDW) based on a large amount of data, we show that our approach outperforms a wide range of state-of-the-art methods. We also deployed our system in the QQ Browser App of Tencent and performed online A/B testing with real users. Results suggest that our approach increased the video click through rate by $23% over implicit-feedback collaborative filtering (IFCF), a scheme available in Apache Spark's MLlib.

Recommendation for user generated content sites has gained significant attention nowadays. To satisfy the niche tastes of users, product recommendation poses more challenges due to the data sparsity issue. This work is motivated by a real world online video recommendation problem, where the click records database suffers from sparseness of video inventory and video tags. Targeting the long tail phenomena of user behavior and sparsity of item features, we propose a personalized compound recommendation framework for online video recommendation called Dirichlet mixture probit model for information scarcity (DPIS). Assuming that each record is generated from a representation of user preferences, DPIS is a probit classifier utilizing record topical clustering on the user part for recommendation. As demonstrated by the real-world application, the proposed DPIS achieves better performance than traditional methods.

Cloud offloading has recently attracted a substantial amount of attention from both industry and academia. This new generation of service, beyond conventional task scheduling and management, utilizes the abundant computation capacity from the public clouds and takes it as a part of external resource on the mobile devices. In this paper, we find that the cloud offloading can potentially increase the running latency of the offloaded tasks. Our measurement shows that the cloud offloading systems, such as Gaikai, can introduce up to 400ms communication latency between the customers and the remote offloading servers. This creates a severe bottleneck to offload the delay sensitive tasks. To mitigate such a challenge, we suggest Cloud Offloading on Customer-Provided Resources, which utilizes the local resources of the customers, such as their home PCs, to minimize the latency during the task offloading. We discuss the framework design based on our commercial system SpotCloud and propose a travel-aware protocol to further address the latency problems when the customers are traveling far from their home PCs. Our model analysis and trace-based simulation indicate that the task execution latency can be reduced by 60% when using the customer-provided offloading service. It is reasonable to believe that such a framework can serve as a useful complement to the conventional cloud offloading on pubic clouds.

Recently, many video sharing sites provide external links so that their video or audio contents can be embedded into external web sites. For example, users can copy the embedded URLs of the videos of YouTube and post the URL links on their own blogs. Clearly, the purpose of such function is to increase the distribution of the videos and the associated advertisement. Does this function fulfill its purpose and what is the quantification? In this paper, we provide a comprehensive measurement study and analysis on these external links to answer these two questions. With the traces collected from two major video sharing sites, YouTube and Youku of China, we show that the external links have various impacts on the popularity of the video sharing sites. More specifically, for videos that have been uploaded for eight months in Youku, around 15% of views can come from external links. Some contents are densely linked. For example, comedy videos can attract more than 800 external links on average. We also study the relationship between the external links and the internal links. We show that there are correlations; for example, if a video is popular itself, it is likely to have a large number of external links. Another observation we find is that the external links usually have a higher impact on Youku than that of YouTube. We conjecture that it is more likely that the external links have higher impact for a regional site than a worldwide site.

In the past few years, we have witnessed a flourish of online social network sites (OSNs). In this kind of websites, the users are not only information consumers, but also actively upload contents of their own to the OSNs. Being sharply different from the conventional sites, OSNs have attracted many studies recently. These studies, however, mainly focus on the social behaviors within OSNs, e.g., the user-user interaction, and the distribution of the time users spend on certain OSNs, etc. There are much fewer studies on the characteristics of the traffic patterns of OSNs upon the Internet infrastructure, even though it is important for the ISPs and the OSNs alike. One major difficulty is that it is not easy to obtain data from the private ISP backbone routers. In this paper, we collect data from two backbone routers of China Telecom and present the traffic patterns of several OSNs from the angle of the network layer. As well, we also compare the OSN traffic with three types of traditional websites like forums, search engines and news websites. We find that the traffic pattern in network layer is quite intuitively similar as what we can see in the application layer: users are the most active in OSNs, and are slightly less active in forums, and they are the least active in News websites. Besides that, we also can see that the traffic pattern in search engines is quite different from the other three.

Recently, many video sharing sites provide external links so that their video or audio contents can be embedded into external web sites. For example, users can copy the embedded URLs of the videos of YouTube and post on their own blogs. Clearly, the purpose of such functionality is to increase the distribution of the videos and the associated advertisement. In this paper, we provide a comprehensive measurement study and analysis on these external links. With the traces collected from two major VOD sites, YouTube and Youku of China, we show that the external links have various impact on the popularity of the VOD sites. More specifically, for videos that have been uploaded for eight months in Youku, around 15% of views can come from external links. Some contents are densely linked, for example, the comedy videos can attract more than 800 external links on average. We also study the relationship between the external links and the internal links. We show that there are correlations; for example, if a video is popular itself, it will likely have a larger number of external links. Another observation is that we always find that the external links of Youku usually have a higher impact than that of YouTube. We conjecture that a more regional site may enjoy a relatively higher impact from the external links.

For the recent two years, YouTube has already become one of the most successful VOD systems. Since it is established in early 2005, the number of the visits to YouTube not only maintain a high level, but also steadily increases even these days. It is widely believed that the online social networks contribute to this success. To study one new aspect of how much social networks contribute to the popularity of videos in YouTube, this paper provides an in-depth study into the effects of the external links in YouTube. Our study shows interesting characteristics of external links in YouTube. The existed studies reveal that `Music' and `Comedy' videos are the most popular ones. Compared with this, our study shows that the `Sports' and `Science & Technology' videos gains the most largest proportion of views from external links. Also, the percentage of views from external links reduces with power law. More specifically, we find that the top-5 external links contribute the popularity of the video most in the first four days and percentage can be up to 25%. For the video over than one years old, the top-5 external links are also able to contribute a stable part at about 4.5%. To end this, we hope our work can serve as a initial step for the study of the external environment.