Text summarization aims to offer a highly condensed and valuable information that expresses the main ideas of the text. Most previous researches focus on extractive models. In this work, we put forward a new generative model based on convolutional seq2seq architecture. A hierarchical CNN framework is much more efficient than the conventional RNN seq2seq models. We also equip our model with a copying mechanism to deal with the rare or unseen words. Additionally, we incorporate a hierarchical attention mechanism to model the keywords and key sentences simultaneously. Finally we verify our model on two real-life datasets, GigaWord and DUC corpus. The experiment results verify the effectiveness of our model as it outperforms state-of-the-art alternatives consistently and statistical significantly.

Fine-grained Named entity recognition (NER) is crucial to natural language processing (NLP) applications like relation extraction and knowledge graph construction. Most existing fine-grained NER systems suffer from inefficiency problem as they use manually annotated training datasets. To address such issue, our NER system could automatically generate datasets from Wikipedia in distant supervision paradigm through mapping hyperlinks in Wikipedia documents to Freebase. In addition, previous NER models can not effectively process fine-grained labels with more than 100 types. So we introduce a ’BIO’ tagging strategy which can identify the position and type attributes simultaneously. Such tagging scheme transfers NER problem into a sequence-to-sequence (seq2seq) based issue. We propose a seq2seq framework to comprehend the input sentence in a comprehensive way. Specifically, we adopt a Bi-LSTM as the encoder to equally process the past and future information of the input. Then we add a self-attention mechanism to handle the long-term dependency problem in a long sequence. When classifying the entity tags, we choose CRF model as it adds more constraints to avoid position logical problem. Experiments are performed on large-scale datasets for fine-grained NER tasks. Experimental results verify the effectiveness of FSeqC, and it outperforms other state-of-the-art alternatives consistently and significantly.

With the rapid growth of patent applications, it has become an urgent problem to automatically classify the accepted patent application documents accurately and quickly. Most previous patent automatic classification studies are based on feature engineering and traditional machine learning methods like SVM, and some even rely on the knowledge of domain experts, hence they suffer from low accuracy problem and have poor generalization ability. In this paper, we propose a patent automatic classification method via the symmetric hierarchical convolution neural network (CNN) named PAC-HCNN. We use the title and abstract of the patent as the input data, and then apply the word embedding technique to segment and vectorize the input data. Then we design a symmetric hierarchical CNN framework to classify the patents based on the word embeddings, which is much more efficient than traditional RNN models dealing with texts, meanwhile keeping the history and future information of the input sequence. We also add gated linear units (GLUs) and residual connection to help realize the deep CNN. Additionally, we equip our model with a self attention mechanism to address the long-term dependency problem. Experiments are performed on large-scale datasets for Chinese short text patent classification. Experimental results prove our proposed model’s effectiveness, and it performs better than other state-of-the-art models significantly and consistently on both fine-grained and coarse-grained classification.

Knowledge representation learning is representing entities and relations in a knowledge graph as dense low-dimensional vectors in the continuous space, which explores the features and properties of the graph. Such a technique can facilitate the computation and reasoning on the knowledge graphs, which benefits many downstream tasks. In order to alleviate the problem of insufficient entity representation learning caused by sparse knowledge graphs, some researchers propose knowledge graph embedding models based on instances and concepts, which utilize the latent semantic connections between concepts and instances contained in the knowledge graphs to enhance the knowledge graph embedding. However, they model instances and concepts in the same space or ignore the transitivity of isA relations, leading to inaccurate embeddings of concepts and instances. To address the above shortcomings, we propose a knowledge graph embedding model that differentiates concepts and instances based on spatial transformation—CIST. The model alleviates the gathering issue of similar instances or concepts in the semantic space by modeling them in different embedding spaces, and adds a learnable parameter to adjust the neighboring range for concept embedding to distinguish hierarchical information of different concepts, thus modeling the transitivity of isA relations. The above features of instances and concepts serve as auxiliary information so that thoroughly modeling them could alleviate the insufficient entity representation learning issue. For the experiments, we chose two tasks, i.e., link prediction and triple classification, and two real-life datasets: YAGO26K-906 and DB111K-174. Compared with state of the arts, CIST achieves an optimal performance in most cases. Specifically, CIST outperforms the SOTA model JOIE by 51.1% on Hits@1 in link prediction and 15.2% on F1 score in triple classification.

Heterogeneous information networks (HINs) are a key resource in many domain-specific retrieval and recommendation scenarios and in conversational environments. Current approaches to mining graph data often rely on abundant supervised information. However, supervised signals for graph learning tend to be scarce for a new task and only a handful of labeled nodes may be available. Meta-learning mechanisms are able to harness prior knowledge that can be adapted to new tasks. In this article, we design meta-learning framework for heterogeneous information networks ( META-HIN ), for few-shot learning problems on HINs. To the best of our knowledge, we are among the first to design a unified framework to realize the few-shot learning of HINs and facilitate different downstream tasks across different domains of graphs. Unlike most previous models, which focus on a single task on a single graph, META-HIN is able to deal with different tasks (node classification, link prediction, and anomaly detection are used as examples) across multiple graphs. Subgraphs are sampled to build the support and query set. Before being processed by the meta-learning module, subgraphs are modeled via a structure module to capture structural features. Then, a heterogeneous Graph Neural Network module is used as the base model to express the features of subgraphs. We also design a Generative Adversarial Network-based contrastive learning module that is able to exploit unsupervised information of the subgraphs. In our experiments, we fuse several datasets from multiple domains to verify META-HIN ’s broad applicability in a multiple-graph scenario. META-HIN consistently and significantly outperforms state-of-the-art alternatives on every task and across all datasets that we consider.

Node importance estimation involves assigning a global importance score to each node in a graph, pivotal to various subsequent tasks, including recommendation, network dismantling, etc. Prior research involves pre-training classification tasks using node labels and structural information, followed by computing node importance scores as a downstream regression task. However, a gap exists caused by the inconsistency between the pre-training and downstream tasks, which tends to exert negative transfer. This paper proposes to narrow down the gap for node importance estimation by implementing a multi-view technique, including node-view for context and graph-view for structure. Specifically, in node-view, we devise soft prompts by encoding node information, which enables the model to capture structural features within a semantic context; afterward, the downstream node regression task is aligned with pre-training by inserting prompt patterns. In graph-view, we introduce virtual nodes, which are learnably inserted based on node importance, to create a prompt graph. High-importance nodes in the original graph are linked to more virtual nodes, improving their embeddings in subsequent propagation steps. Such enhancement increases their importance scores in downstream tasks, improving the model's ability to distinguish significant nodes effectively. Additionally, the prompts from different views are fused through multi-view contrastive learning to further enhance the expressiveness of the node embeddings. We empirically evaluate our model on four public datasets, which are shown to outperform other state-of-the-art alternatives significantly and consistently.