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.

As a vital measure to obtain valuable information and intelligence, web news are flooding all corners of the Internet anytime, anywhere. Traditionally, templates or hand-designed features are utilized to extract the content from web pages, but these models have higher time cost and lower extensibility. Recently, many scholars leverage DOM-tree-based or text-density-based models to extract the contents which have better extensibility and lower time cost, but most of them are hard to extract the content accurately and completely and are easy to introduce the noises. In this paper, we propose a title-based web content extracting model TWCEM to extract the contents of each web page, which leverage the title information to extract the web content. Compared with other extraction model, TWCEM can filter the noises effectively and locate the content positions more accurately. In experiment, we evaluate the proposed model on real-life websites, and TWCEM achieves state-of-the-art results and outperforms its competitors on both extraction performance and time cost.

Scene text recognition is a frontier research direction in optical character recognition (OCR) technology. Considering that the compatibility of the existing CRNN method under the PaddlePaddle framework is not friendly, and the recognition accuracy needs to be improved. Therefore, this paper proposes an efficient RBC model constructed by fusing ResNet34 and Bi-LSTM backbone network and Connectionist Temporal Classification (CTC) transcription method to solve the problem of scene text recognition. By coding based on the PaddlePaddle framework and experimenting with shared dataset, the results show that the RBC model is well compatible with the PaddlePaddle framework, and the average accuracy has reached 85.91%, which is 1.8% higher than the optimal baseline model. The data and code of this article, please link to the URL https://aistudio.baidu.com/aistudio/projectdetail/1911502?=1622688767906.

Chinese text classification is an important task in data mining, which extracts category features from unstructured contents. Conventional Chinese text classification models only leverage the surface features in the original text, which omits the potential extensional knowledge of each word. To capture the semantic features of each word more comprehensive, this paper proposed a Chinese news text classification algorithm based on an online knowledge extension and convolutional neural network (OKE-CNN), which leverages both knowledge graph to extend latent semantic information and CNN to obtain the category. Compared with other baseline methods, OKE-CNN can utilize the surface and latent features, simultaneously, which can be adapted to complex scenes, e.g., sparse data and unclear topics. In our experiment, OKE-CNN exhibits superior performance and achieves 97.94% and 87.03% on THUCNews and TouTiao datasets, separately, over SOTA competitors.