Authentication email has already been sent, please check your email box: and activate it as soon as possible.
You can login to My Profile and manage your email alerts.
If you haven’t received the email, please:
|
|
There are 46 papers published in subject: > since this site started. |
Select Subject |
Select/Unselect all | For Selected Papers |
Saved Papers
Please enter a name for this paper to be shown in your personalized Saved Papers list
|
1. A BERT based Multi-task Learning Model for Judgment Prediction | |||
Yang Ze,Zhang Lei | |||
Computer Science and Technology 25 March 2019 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Judgment prediction is a difficult problem in judicial field. Given a fact description, judge need to conduct several documents to decide the related articles. This task is complex and requires a lot of energy. Previous works always treat this task as a multi-label learning paradigm for judgment prediction. These work usually neglect the external knowledge, thus the performance may be limited. In this paper, this topic use a multi-task learning framework with pretrained external knowledge to address this issues, and this topic propose a BERT based multi-task learning model(BMM for short). Specially, BMM use a pretrained BERT model to obtain external knowledge, then a multi-task learning framework is incorporated to learn multi-label classification and language model jointly. Experimental results on three real-world datasets demonstrate that the proposed model achieves significant improvements over state-of-the-art methods. | |||
TO cite this article:Yang Ze,Zhang Lei. A BERT based Multi-task Learning Model for Judgment Prediction[OL].[25 March 2019] http://en.paper.edu.cn/en_releasepaper/content/4747962 |
2. Sentence Encoding: attention with another self | |||
Zhang Xinnan,Li Wei | |||
Computer Science and Technology 16 January 2019 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:In this paper, we propose an attention based structure to do sentence encoding in natural language inference(NLI) task. This attention strcture do soft-alignment between word features and character features, encoding both of them to a joint feature vector space, it helps models handle more unknown words\' information, enhance the model performance in some specific situations. Our experiments show that this structure work well in Stanford Natural Language Inference(SNLI) dataset and The Multi-Genre Natural Language Inference(MultiNLI) corpus. | |||
TO cite this article:Zhang Xinnan,Li Wei. Sentence Encoding: attention with another self[OL].[16 January 2019] http://en.paper.edu.cn/en_releasepaper/content/4747072 |
3. AJWE: Jointly Learning Chinese Word Embeddings with Heterogeneous Attention | |||
Liu Jie,Wang Yulong | |||
Computer Science and Technology 09 January 2019 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Much attention has been drawn to leveraging the sub-word information to improve word representation, especially in some morphological language like Chinese. Previous studies on Chinese word embeddings has explored diverse fine-grained sub-word information, such as character, radical, component, stoke n-grams. However, all of them do not distinguish the semantic contribution of word to the context and are weak at handling the ambiguity of characters and sub-character components as well. In this paper, we propose AJWE, a jointly model for learning Chinese word embeddings with heterogeneous attention. we explore an external self-attention mechanism to learning the word semantic contribution to the context, specially propose a bias-attention approach for in-ternal sub-word morphemes to address the ambiguity issue. Evaluation on the word similarity, word analogy, text classification and name entity recognition demonstrates that our model outperforms existing state-of-the-art methods. | |||
TO cite this article:Liu Jie,Wang Yulong. AJWE: Jointly Learning Chinese Word Embeddings with Heterogeneous Attention[OL].[ 9 January 2019] http://en.paper.edu.cn/en_releasepaper/content/4746998 |
4. Bug Severity Prediction Based on GRU and Various Features | |||
Peng Li,Zhou Ying | |||
Computer Science and Technology 12 April 2018 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:With the increasing of scale and complexity of the software, the defect in software become unavoidable. Therefore, bug fixing is an essential activity in software maintenance. The severity of a reported bug is a critical factor in deciding how soon it needs to be fixed. Unfortunately, although there are some rules available on how to justify the level of severity for a given bug, the manual process remains a big problem which can increase fixing time and achieve low precision. To address this issue, we propose GRUModel, a novel model based on neural network that achieve good performance in both prediction accuracy and fixing time. Specifically, we utilize various features not only bug description as input data (e.g., component, developer, priority and severity). To evaluate our approach, we measured the effectiveness of our study by using about 180,000 golden bug reports extracted from five open source products (platform,cdt,jdt,pde and birt). The experiment results demonstrate that our approach predict the severity with a higher accuracy (both precision and recall vary between 0.72-0.85), compared with the existing methods such as Naive Bayes. | |||
TO cite this article:Peng Li,Zhou Ying. Bug Severity Prediction Based on GRU and Various Features[OL].[12 April 2018] http://en.paper.edu.cn/en_releasepaper/content/4744487 |
5. Sentence Similarity measurement Based on the combination of Siamese recurrent network model and Word Alignment model | |||
ZHAO Yinge,XU Weiran | |||
Computer Science and Technology 02 December 2017 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:In this paper,we introduce a new method to compute the sentences pair similarity which is the combination of a Siamese recurrent neural network model and a word alignment model. In our Siamese recurrent neural network model, we use the original sentence information and interaction information as the input of our network with Google pre-trained word2vec, and calculate the cosine value of the Siamese network outputs as a feature of the sentence pair similarity. While in the alignment model, it calculates the number of alignment word pairs and regards the ration as the sentences pair similarity. In the last we combine the two features in a simple way. In the STS2016 SemEval test data, out model get the state of art result in the news headlines section and the top 2 in the median result. | |||
TO cite this article:ZHAO Yinge,XU Weiran. Sentence Similarity measurement Based on the combination of Siamese recurrent network model and Word Alignment model[OL].[ 2 December 2017] http://en.paper.edu.cn/en_releasepaper/content/4742223 |
6. Synonymous Entity Recognition based on Feature Fusion | |||
Cai Desheng,He Jingjing,Wu Gongqing,Xuegang Hu | |||
Computer Science and Technology 26 April 2017 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:It is necessary to unify data from different sources in Web information fusion as the same entity has different expressions in the real world. People usually use the information obtained from the search engine to make decisions in daily life. The idea of decision-making by means of search engines' information is used for synonymous entity recognition. In this paper, we propose a method of synonymous entity recognition based on feature fusion, which used multiplicative information fusion technique to fuse numbers of results of synonymous entity recognition. In order to utilize the information obtained from the search engine, we combine entity name and other entity features as query words and design a new similarity function VarSim to measure entities' similarity. The F-score of algorithm of based on feature fusion on the two datasets are 13.42% and 1.81% higher than that based on the entity name, which demonstrates the effectiveness of the synonymous entity recognition approach based on feature fusion. | |||
TO cite this article:Cai Desheng,He Jingjing,Wu Gongqing, et al. Synonymous Entity Recognition based on Feature Fusion[OL].[26 April 2017] http://en.paper.edu.cn/en_releasepaper/content/4727733 |
7. Enterprise Entity Extraction in News with Knowledge Base | |||
Zhang Kuai,Xin Xin | |||
Computer Science and Technology 24 April 2017 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:We focus on the issue of identify the name of enterprise in news text. The previous method only utilizes the textual information to extract the entities, in this paper, we introduce an enterprise knowledge base to generate short names of companies by man-made rules, then we filter the wrong candidates by their feature on the corpus. The experiment results show our mehtod which combiene the semantic feature and knowledge base information achieves a good performance on our news dataset. | |||
TO cite this article:Zhang Kuai,Xin Xin. Enterprise Entity Extraction in News with Knowledge Base[OL].[24 April 2017] http://en.paper.edu.cn/en_releasepaper/content/4728966 |
8. Fine-grained Coordinated Cross-lingual Text Stream Alignment for Endless Language Knowledge Acquisition | |||
GE Tao, DOU Qing, JI Heng, CUI Lei, CHANG Baobao, PAN Xiaoman, SUI Zhifang, ZHOU Ming | |||
Computer Science and Technology 24 April 2017 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:This paper proposes to study fine-grained coordinated cross-lingual text stream alignment through a novel information network decipherment paradigm. We use Burst Information Networks as media to represent the text streams and present a simple yet effective network decipherment algorithm with diverse clues to decipher the networks for accurate text stream alignment. Extensive experiments on Chinese-English coordinated news streams show our approach can harvest high-quality alignments from large amounts of streaming data for endless language knowledge mining. | |||
TO cite this article:GE Tao, DOU Qing, JI Heng, et al. Fine-grained Coordinated Cross-lingual Text Stream Alignment for Endless Language Knowledge Acquisition[OL].[24 April 2017] http://en.paper.edu.cn/en_releasepaper/content/4726125 |
9. Improving Social Media Text Summarization by Learning Sentence Weight Distribution | |||
Xu Jing-Jing, Sun Xu, Ren Xuan-Cheng | |||
Computer Science and Technology 24 April 2017 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract:Recently, encoder-decoder models are widely used in social media text summarization. However, these models sometimes select noise words in irrelevant sentences as part of a summary by error, thus declining the performance. In order to inhibit irrelevant sentences and focus on key information, we propose an effective approach by learning sentence weight distribution. In our model, we build a multi-layer perceptron to predict sentence weights. During training, we use the ROUGE score as an alternative to the estimated sentence weight, and try to minimize the gap between estimated weights and predicted weights. In this way, we encourage our model to focus on the key sentences, which have high relevance with the summary. Experimental results show that our approach outperforms baselines on a large-scale social media corpus. | |||
TO cite this article:Xu Jing-Jing, Sun Xu, Ren Xuan-Cheng. Improving Social Media Text Summarization by Learning Sentence Weight Distribution[OL].[24 April 2017] http://en.paper.edu.cn/en_releasepaper/content/4726405 |
10. Towards A Noise-Tolerant Neural Network Model for Distant Supervised Relation Extraction | |||
JIANG Ting-Song, SUI Zhi-Fang | |||
Computer Science and Technology 20 April 2017 | |||
Show/Hide Abstract | Cite this paper︱Full-text: PDF (0 B) | |||
Abstract: Distantly supervised relation extraction has been widely used to extract semantic relations from text. However, it suffers from wrong labeling problems and hinders the performance of a model trained on such noisy data. To deal with this problem, previous neural network model assumed at least one instance is true and only selected the most likely one instance in a bag for training, which missed rich information by discarding all other true positive instances. Instead of using the at-least-one assumption, we assume that most of the distantly labeled instances are true positive instances and true positive instances often share the same feature patterns but false positive instances do not. We argue that all instances should be given to the model and leave the model to decide the contribution of different instances. Therefore we propose two kinds of methods to dynamically determine the weights of all the instances and make the neural network model more tolerant to noise. Experiments show that our approach is effective, and it outperforms several competitive baseline methods. | |||
TO cite this article:JIANG Ting-Song, SUI Zhi-Fang. Towards A Noise-Tolerant Neural Network Model for Distant Supervised Relation Extraction[OL].[20 April 2017] http://en.paper.edu.cn/en_releasepaper/content/4726134 |
Select/Unselect all | For Selected Papers |
Saved Papers
Please enter a name for this paper to be shown in your personalized Saved Papers list
|
About Sciencepaper Online | Privacy Policy | Terms & Conditions | Contact Us
© 2003-2012 Sciencepaper Online. unless otherwise stated