NEWS ARTICLE TEXT CLASSIFICATION AND SUMMARY FOR AUTHORS AND TOPICS
Aviel J. Stein1, Janith Weerasinghe2Spiros Mancoridis11 and Rachel Greenstadt2 1Drexel University, Philadelphia, Pennsylvania, USA 2,3New York University, New York, USA
News articles are important for providing timely, historic information. However, the Internet is replete with text that may contain irrelevant or unhelpful information, therefore means of processing it and distilling content is important and useful to human readers as well as information extracting tools. Some common questions we may want to answer are “what is this article about?” and “who wrote it?”. In this work we compare machine learning models for evaluating two common NLP tasks, topic and authorship attribution, on the 2017 Vox Media dataset. Additionally, we use the models to classify on a subsection, about ~20%, of the original text which show to be better for classification than the provided blurbs. Because of the large number of topics, we take into account topic overlap and address it via top-n accuracy and hierarchical groupings of topics. We also consider edge cases in authorship by classifying on inter-topic and intra-topic author distributions. Our results show that both topics and authors readily identifiable consistently perform best when using neural networks rather than support vector, random forests, or naive Bayes classifiers, although the latter methods perform acceptably.
Natural Language Processing, Topic Classification, Author Attribution, Summarization, Machine Learning
Full Paper
https://aircconline.com/csit/papers/vol10/csit101401.pdf
Volume Link :
http://airccse.org/csit/V10N14.html
CADA-FVAE-GAN: ADVERSARIAL TRAINING FOR FEW-SHOT EVENT DETECTION
Xiaoxiang Zhu, Mengshu Hou, Xiaoyang Zeng and Hao Zhu University of Electronic Science and Technology of China, Chengdu, China
Most supervised systems of event detection (ED) task reply heavily on manual annotations and suffer from high-cost human effort when applied to new event types. To tackle this general problem, we turn our attention to few-shot learning (FSL). As a typical solution to FSL, cross- modal feature generation based frameworks achieve promising performance on images classification, which inspires us to advance this approach to ED task. In this work, we propose a model which extracts latent semantic features from event mentions, type structures and type names, then these three modalities are mapped into a shared low-dimension latent space by modality-specific aligned variational autoencoder enhanced by adversarial training. We evaluate the quality of our latent representations by training a CNN classifier to perform ED task. Experiments conducted on ACE2005 dataset show an improvement with 12.67% on F1-score when introducing adversarial training to VAE model, and our method is comparable with existing transfer learning framework for ED.
Event Detection, Few-Shot Learning, Cross-modal generation, Variational autoencoder, GAN
For More Details :
https://aircconline.com/csit/papers/vol10/csit101402.pdf
Volume Link :
http://airccse.org/csit/V10N14.html
SEMANTIC MANAGEMENT OF ENTERPRISE INFORMATION SYSTEMS THROUGH ONTOLOGIES
Valentina Casola and Rosario Catelli University of Naples Federico II, Naples, Italy
This article introduces a model for cloud-aware enterprise governance with a focus on its semantic aspects. It considers the need for Business-IT/OT and Governance-Security alignments. The proposed model suggests the usage of ontologies as specific tools to address the governance of each IT/OT environment in a holistic way. The concrete utilization of ISO and NIST standards allows to correctly structure the ontological model: in fact, by using these wellknown international standards it is possible to significantly reduce terminological and conceptual inconsistencies already in the design phase of the ontology. This also brings a considerable advantage in the operational management phase of the company certifications, congruently aligned with the knowledge structured in this manner. The semantic support within the model suggests further possible applications in different departments of the company, with the aim of evaluating and managing projects in an optimal way, integrating different but important points of view of stakeholders.
Cloud, Enterprise, Governance, Information management, Ontology, Semantic systems
For More Details :
https://aircconline.com/csit/papers/vol10/csit101403.pdf
Volume Link :
http://airccse.org/csit/V10N14.html
INJECTING EVENT KNOWLEDGE INTO PRE- TRAINED LANGUAGE MODELS FOR EVENT EXTRACTION
Zining Yang1, Siyu Zhan1,Mengshu Hou1,Xiaoyang Zeng1and Hao Zhu2 1 1,2University of Electronic Science & Technology of China, Chengdu, China
The recent pre-trained language model has made great success in many NLP tasks. In this paper, we propose an event extraction system based on the novel pre-trained language model BERT to extract both event trigger and argument. As a deep-learning based method, the size of the training dataset has a crucial impact on performance. To address the lacking training data problem for event extraction, we further train the pretrained language model with a carefully constructed in-domain corpus to inject event knowledge to our event extraction system with minimal efforts. Empirical evaluation on the ACE2005 dataset shows that injecting event knowledge can significantly improve the performance of event extraction.
Natural Language Processing, Event Extraction, BERT, Lacking Training Data Problem
For More Details :
https://aircconline.com/csit/papers/vol10/csit101404.pdf
Volume Link :
http://airccse.org/csit/V10N14.html
ARABIC LOCATION NAME ANNOTATIONS AND APPLICATIONS
Omar ASBAYOU Lumière University, CRTT, Lyon 2, France
This paper show how location named entity (LNE) extraction and annotation, which makes part of our named entity recognition (NER) systems, is an important task in managing the great amount of data. In this paper, we try to explain our linguistic approach in our rule-based LNE recognition and classification system based on syntactico-semantic patterns. To reach good results, we have taken into account morpho-syntactic information provided by morpho-syntactic analysis based on DIINAR database, and syntactico-semantic classification of both location name trigger words (TW) and extensions. Formally, different trigger word sense implies different syntactic entity structures. We also show the semantic data that our LNE recognition and classification system can provide to both information extraction (IE) and information retrieval(IR).The XML database output of the LNE system constituted an important resource for IE and IR. Future project will improve this processing output in order to exploit it in computerassisted Translation (CAT).
Location name annotations, Location named entities, Information retrieval, Information extraction
For More Details :
https://aircconline.com/csit/papers/vol10/csit101405.pdf
Volume Link :
http://airccse.org/csit/V10N14.html
JOINT EXTRACTION OF ENTITY AND RELATION WITH INFORMATION REDUNDANCY ELIMINATION
Yuanhao Shen and Jungang Han Xi`an University of Posts and Telecommunications, Xi`an, China
To solve the problem of redundant information and overlapping relations of the entity and relation extraction model, we propose a joint extraction model. This model can directly extract multiple pairs of related entities without generating unrelated redundant information. We also propose a recurrent neural network named Encoder-LSTM that enhances the ability of recurrent units to model sentences. Specifically, the joint model includes three sub-modules: the Named Entity Recognition sub-module consisted of a pre-trained language model and an LSTM decoder layer, the Entity Pair Extraction sub-module which uses Encoder-LSTM network to model the order relationship between related entity pairs, and the Relation Classification submodule including Attention mechanism. We conducted experiments on the public datasets ADE and CoNLL04 to evaluate the effectiveness of our model. The results show that the proposed model achieves good performance in the task of entity and relation extraction and can greatly reduce the amount of redundant information.
Joint Model, Entity Pair Extraction, Named Entity Recognition, Relation Classification, Information Redundancy Elimination.
For More Details :
https://aircconline.com/csit/papers/vol10/csit101406.pdf
Volume Link :
http://airccse.org/csit/V10N14.html
ISOLATING WORD LEDOMAIN-TRANSFERABLE METHOD FOR NAMED ENTITY RECOGNITION TASK
Vladislav Mikhailov1,2and Tatiana Shavrina1,2 1Sberbank, Moscow, Russia 2Higher School of Economics, Moscow, Russia
Named Entity Recognition (NER) is a fundamental task in the fields of natural language processing and information extraction. NER has been widely used as a standalone tool or an essential component in a variety of applications such as question answering, dialogue assistants and knowledge graphs development. However, training reliable NER models requires a large amount of labelled data which is expensive to obtain, particularly in specialized domains. This paper describes a method to learn a domain- specific NER model for an arbitrary set of named entities when domain-specific supervision is not available. We assume that the supervision can be obtained with no human effort, and neural models can learn from each other. The code, data and models are publicly available.
Named Entity Recognition, BERT-based Models, Russian Language
For More Details :
https://aircconline.com/csit/papers/vol10/csit101407.pdf
Volume Link :
http://airccse.org/csit/V10N14.html
CHINESE MEDICAL QUESTION ANSWER MATCHING BASED ON INTERACTIVE SENTENCE REPRESENTATION LEARNING
Xiongtao Cui and Jungang Han Xi’an University of Posts and Telecommunications, Xi’an, China
Chinese medical question-answer matching is more challenging than the open-domain questionanswer matching in English. Even though the deep learning method has performed well in improving the performance of question-answer matching, these methods only focus on the semantic information inside sentences, while ignoring the semantic association between questions and answers, thus resulting in performance deficits. In this paper, we design a series of interactive sentence representation learning models to tackle this problem. To better adapt to Chinese medical question-answer matching and take the advantages of different neural network structures, we propose the Crossed BERT network to extract the deep semantic information inside the sentence and the semantic association between question and answer, and then combine with the multi-scale CNNs network or BiGRU network to take the advantage of different structure of neural networks to learn more semantic features into the sentence representation. The experiments on the cMedQA V2.0 and cMedQA V1.0 dataset show that our model significantly outperforms all the existing state-of-the-art models of Chinese medical question answer matching.
Question answer matching, Chinese medical field, interactive sentence representation, deep learning
For More Details :
https://aircconline.com/csit/papers/vol10/csit101408.pdf
Volume Link :
http://airccse.org/csit/V10N14.html
A PATTERN-MINING DRIVEN STUDY ON DIFFERENCES OF NEWSPAPERS IN EXPRESSING TEMPORAL INFORMATION
Yingxue Fu1,2and Elaine Uí Dhonnchadha2 1University of St Andrews, Scotland, UK 2Center for Language and Communication Studies, Trinity College Dublin
This paper studies the differences between different types of newspapers in expressing temporal information, which is a topic that has not received much attention. Techniques from the fields of temporal processing and pattern mining are employed to investigate this topic. First, a corpus annotated with temporal information is created by the author. Then, sequences of temporal information tags mixed with part-of-speech tags are extracted from the corpus. The TKS algorithm is used to mine skip-gram patterns from the sequences. With these patterns, the signatures of the four newspapers are obtained. In order to make the signatures uniquely characterize the newspapers, we revise the signatures by removing reference patterns. Through examining the number of patterns in the signatures and revised signatures, the proportion of patterns containing temporal information tags and the specific patterns containing temporal information tags, it is found that newspapers differ in ways of expressing temporal information.
Pattern Mining, TKS algorithm, Temporal Annotation, Tabloids and Broadsheets
For More Details :
https://aircconline.com/csit/papers/vol10/csit101409.pdf
Volume Link :
http://aircconline.com/csit/abstract/v10n14/csit101409.html
MULTI-LAYER ATTENTION APPROACH FOR ASPECT BASED SENTIMENT ANALYSIS
Xinzhi Ai1,Xiaoge Li1Feixiong Hu2Shuting Zhi1and Likun Hu2 1Xi’an University of Posts and Telecommunications, Xi’an, China, 710121 2Tencent Technology (Shenzhen) Co., Ltd, China, 518057
Based on the aspect-level sentiment analysis is typical of fine-grained emotional classification that assigns sentiment polarity for each of the aspects in a review. For better handle the emotion classification task, this paper put forward a new model which apply Long Short-Term Memory network combine multiple attention with aspect context. Where multiple attention mechanism (i.e., location attention, content attention and class attention) refers to takes the factors of context location, content semantics and class balancing into consideration. Therefore, the proposed model can adaptively integrate location and semantic information between the aspect targets and their contexts into sentimental features, and overcome the model data variance introduced by the imbalanced training dataset. In addition, the aspect context is encoded on both sides of the aspect target, so as to enhance the ability of the model to capture semantic information. The Multi- Attention mechanism (MATT) and Aspect Context (AC) allow our model to perform better when facing reviews with more complicated structures. The result of this experiment indicate that the accuracy of the new model is up to 80.6% and 75.1% for two datasets in SemEval-2014 Task 4 respectively, While the accuracy of the data set on twitter 71.1%, and 81.6% for the Chinese automotive-domain dataset. Compared with some previous models for sentiment analysis, our model shows a higher accuracy.
Aspect-level sentiment analysis, Multiple attention mechanism, LSTM neural network
For More Details :
https://aircconline.com/csit/papers/vol10/csit101410.pdf
Volume Link :
http://aircconline.com/csit/abstract/v10n14/csit101409.html
A TOPOLOGICAL METHOD FOR COMPARING DOCUMENT SEMANTICS
Yuqi Kong1,Fanchao Meng1and Ben Carterette 1University of Delaware, Newark, USA 2Spotify, Greenwich Street, New York, USA
Comparing document semantics is one of the toughest tasks in both Natural Language Processing and Information Retrieval. To date, on one hand, the tools for this task are still rare. On the other hand, most relevant methods are devised from the statistic or the vector space model perspectives but nearly none from a topological perspective. In this paper, we hope to make a different sound. A novel algorithm based on topological persistence for comparing semantics similarity between two documents is proposed. Our experiments are conducted on a document dataset with human judges’ results. A collection of state-of-the-art methods are selected for comparison. The experimental results show that our algorithm can produce highly human-consistent results, and also beats most state-of-the-art methods though ties with NLTK.
Topological Graph, Document Semantics Comparison, Natural Language Processing, Information Retrieval, Topological Persistence
For More Details :
https://aircconline.com/csit/papers/vol10/csit101411.pdf
Volume Link :
http://aircconline.com/csit/abstract/v10n14/csit101409.html
LOCAL SELF-ATTENTION BASED CONNECTIONIST TEMPORAL CLASSIFICATION FOR SPEECH RECOGNITION/p>
Deng Huizhen and Zhang zhaogong Heilongjiang University of China, China
Connectionist temporal classification (CTC) has been successfully applied to end-to-end speech recognition tasks, but its main body recurrent neural network makes parallelization very difficult. Since the attention mechanism has shown very good performance on a series of tasks such as machine translation, handwriting synthesis, and image caption generation for loop sequence generators conditioned on input data. This paper applies the attention mechanism to CTC, and proposes a connectionist temporal classification based on the local self-attention mechanism, in which the cyclic neural network module in the traditional CTC model is replaced by the self- attention module. It shows that it is attractive and competitive in end-to-end speech recognition. The proposed mechanism is based on local self-attention, which uses a sliding mechanism to obtain acoustic features locally. This mechanism effectively models long-term scenarios by stacking multiple sliders to obtain a larger receiving field to achieve online decoding. Moreover, the CTC training joint cross-entropy criterion makes the model converge better. We have completed experiments on the AISHELL-1 dataset. The experiments show that the basic model has a lower character error rate than the existing state-of-the-art models, and the model after cross entropy has been further improved.
Connectionist temporal classification, self-attention mechanism, cross entropy, Speech Recognition
For More Details :
https://aircconline.com/csit/papers/vol10/csit101122.pdf
Volume Link :
http://airccse.org/csit/V10N11.html
APPLICABILITY OF DEEP NEURAL NETWORKS ON THE TASK OF DOCUMENT RETRIEVAL
M. Shoaib Malik1,2and Dagmar Waltemath1 1University Medicine Greifswald, Germany 1Air University, Islamabad, Pakistan
A Deep Neural Network (DNN) can be used to learn higher-level and more abstract representations of a particular input. DNNs have successfully been applied to analysis tasks including image processing, unsupervised feature learning, and natural language processing. DNNs furthermore can improve computing performance when compared to shallower networks, for example in pattern recognition tasks in machine learning. Recent usage of DNNs in search engines for the Web have impacted that technology in industrial scale applications. One example for such an application is deepgif - a search engine for Graphics Interchange Format (GIF) images that is based on a convolutional neural network and takes natural language text as query. In this study, we developed a tool and compared the performance of feed-forward neural networks and deep architectures of recurrent neural network using the case of document retrieval. This study first discusses two architectural setups used to build the models and then provide a detailed comparison of their performance. The goal is to identify the architecture that is most suited for the task of document retrieval.
Deep Neural Network, Machine Learning, Document Retrieval, Feed-Forward Neural Network, Recurrent Neural Network
For More Details :
https://aircconline.com/csit/papers/vol10/csit101123.pdf
Volume Link :
http://airccse.org/csit/V10N11.html
THE DESIGN AND IMPLEMENTATION OF LANGUAGE LEARNING CHATBOT WITH XAI USING ONTOLOGY AND TRANSFER LEARNING
Nuobei SHI, Qin Zeng and Raymond Lee University-Hong Kong Baptist University United International College, Zhuhai, China
In this paper, we proposed a transfer learning-based English language learning chatbot, whose output generated by GPT-2 can be explained by corresponding ontology graph rooted by finetuning dataset. We design three levels for systematically English learning, including phonetics level for speech recognition and pronunciation correction, semantic level for specific domain conversation, and the simulation of “free-style conversation” in English - the highest level of language chatbot communication as ‘free-style conversation agent’. For academic contribution, we implement the ontology graph to explain the performance of free-style conversation, following the concept of XAI (Explainable Artificial Intelligence) to visualize the connections of neural network in bionics, and explain the output sentence from language model. From implementation perspective, our Language Learning agent integrated the mini-program in WeChat as front-end, and fine-tuned GPT-2 model of transfer learning as back-end to interpret the responses by ontology graph. All of our source codes have uploaded to GitHub: https://github.com/p930203110/EnglishLanguageRobot.
NLP-based Chatbot, Explainable Artificial Intelligence (XAI), Ontology graph, GPT-2, Transfer Learning
For More Details :
https://aircconline.com/csit/papers/vol10/csit101124.pdf
Volume Link :
http://airccse.org/csit/V10N11.html
TEXT-BASED EMOTION AWARE RECOMMENDER
John Kalung Leung1,Igor Griva2and William G. Kennedy3 1George Mason University, 4400 University Drive, Fairfax, Virginia 22030, USA 2University,4400 University Drive, Fairfax, Virginia 22030, USA 3George Mason University, 4400 University Drive, Fairfax, Virginia 22030, USA
We apply the concept of users' emotion vectors (UVECs) and movies' emotion vectors (MVECs) as building components of Emotion Aware Recommender System. We built a comparative platform that consists of five recommenders based on content-based and collaborative filtering algorithms. We employed a Tweets Affective Classifier to classify movies' emotion profiles through movie overviews. We construct MVECs from the movie emotion profiles. We track users' movie watching history to formulate UVECs by taking the average of all the MVECs from all the movies a user has watched. With the MVECs, we built an Emotion Aware Recommender as one of the comparative platforms' algorithms. We evaluated the top-N recommendation lists generated by these Recommenders and found the top-N list of Emotion Aware Recommender showed serendipity recommendations.
Context-Aware, Emotion Text Mining, Affective Computing, Recommender Systems, Machine Learning
For More Details :
https://aircconline.com/csit/papers/vol10/csit101009.pdf
Volume Link :
http://airccse.org/csit/V10N10.html
USING HOLOGRAPHICALLY COMPRESSED EMBEDDINGS IN QUESTION ANSWERING
Salvador E. Barbosa Middle Tennessee State University, Murfreesboro, TN, USA
Word vector representations are central to deep learning natural language processing models. Many forms of these vectors, known as embeddings, exist, including word2vec and GloVe. Embeddings are trained on large corpora and learn the word’s usage in context, capturing the semantic relationship between words. However, the semantics from such training are at the level of distinct words (known as word types), and can be ambiguous when, for example, a word type can be either a noun or a verb. In question answering, parts-of-speech and named entity types are important, but encoding these attributes in neural models expands the size of the input. This research employs holographic compression of pre-trained embeddings, to represent a token, its part-of-speech, and named entity type, in the same dimension as representing only the token. The implementation, in a modified question answering recurrent deep learning network, shows that semantic relationships are preserved, and yields strong performance.
Question Answering, Vector Embeddings, Holographic Reduced Representations, DrQA, SQuAD
For More Details :
https://aircconline.com/csit/papers/vol10/csit100919.pdf
Volume Link :
http://airccse.org/csit/V10N09.html
A Semantic Question Answering in a Restricted Smart Factory Domain Attaching to Various Data Sources
Orçun Oruç Technische Universität Dresden, Software Technology Group, Nöthnitzer Strasse 46, 01187, Dresden
Industrial manufacturing has become more interconnected between smart devices such as the industry of things edge devices, tablets, manufacturing equipment, and smartphones. Smart factories have emerged and evolved with digital technologies and data science in manufacturing systems over the past few years. Smart factories make complex data enables digital manufacturing and smart supply chain management and enhanced assembly line control. Nowadays, smart factories produce a large amount of data that needs to be apprehensible by human operators and experts in decision making. However, linked data is still hard to understand and interpret for human operators, thus we need a translating system from linked data to natural language or summarization of the volume of linked data by eliminating undesired results in the linked data repository. In this study, we propose a semantic question answering in a restricted smart factory domain attaching to various data sources. In the end, we will perform qualitative and quantitative evaluation of the semantic question answering, as well as discuss findings and conclude the main points with regard to our research questions.
Semantic Web, Web 3.0, Information Retrieval, Natural Language Processing, Industry 4.0.
For More Details :
https://aircconline.com/csit/papers/vol10/csit100920.pdf
Volume Link :
http://airccse.org/csit/V10N09.html
Journals by Area
Conference Publicity
To list your conference in this page, please contact us
