Cross-Domain Sentiment Classification via a Bifurcated-LSTM

Ji, Jinlong; Luo, Changqing; Chen, Xuhui; Yu, Lixing; Pan, Li

doi:10.1007/978-3-319-93034-3_54

Cited by 15 publications

(10 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Deep learning models constitute another important group of techniques for cross-domain sentiment classification. Various neural network architectures are exploited to compute the representation vectors for reviews, such as, stacked auto-encoders (SAE) [33], [34], fully connected neural networks [21], convolutional neural networks (CNN) [25], and also recurrent neural networks (RNN) typically with long short-term memory (LSTM) units [23]. To enhance the expressive power of neural networks, there are various works investigating the use of recently developed attention and memory mechanisms [22], [24].…”

Section: Related Workmentioning

confidence: 99%

“…Comparing the domain adaption strategies used by various state-of-the-art techniques, we can see that, in addition to the main task of sentiment classification, they usually enhance their learning through preparing extra tasks like detecting whether a pivot co-occurs with a domain-specific word, whether the different versions of the same pivot word in different domains possess similar enough representation vectors, whether a review contains a pivot, or whether the reviews from the source and target domains can be distinguished in the representation space, and so on; we refer to these as the auxiliary learning tasks. The learning algorithms are mostly built on spectral approaches which explore and preserve inherent data structure through matrix decompositions [10], [11], [17], or neural networks which directly learn the review representations through structured processing of the content words based on different network architectures and exhaustive training [21], [23]- [25].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Cross-Domain Sentiment Encoding through Stochastic Word Embedding

Hao

Hong

et al. 2020

IEEE Trans. Knowl. Data Eng.

View full text Add to dashboard Cite

Sentiment analysis is an important topic concerning identification of feelings, attitudes, emotions and opinions from text. To automate such analysis, a large amount of example text needs to be manually annotated for model training. This is laborious and expensive, but the cross-domain technique is a key solution to reducing the cost by reusing annotated reviews across domains. However, its success largely relies on the learning of a robust common representation space across domains. In the recent years, significant effort has been invested to improve the cross-domain representation learning by designing increasingly more complex and elaborate model inputs and architectures. We support that it is not necessary to increase design complexity as this inevitably consumes more time in model training. Instead, we propose to explore the word polarity and occurrence information through a simple mapping and encode such information more accurately whilst managing lower computational costs. The proposed approach is unique and takes advantage of the stochastic embedding technique to tackle cross-domain sentiment alignment. Its effectiveness is benchmarked with over ten data tasks constructed from two review corpora and it is compared against ten classical and state-of-the-art methods.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Cross-Domain Sentiment Encoding through Stochastic Word Embedding

Hao

Hong

et al. 2020

IEEE Trans. Knowl. Data Eng.

View full text Add to dashboard Cite

show abstract

“…Deep learning techniques, such as Artificial Neural Network (ANN), Recurrent Neural Network (RNN), LSTM, have been widely employed in various applications due to their strong capability in data pattern and correlation mining [34], [35]. In TrafficChain, it is obvious that the road passing time cost for and the number of vehicles on each road segment are highly related to the history data on the same road segment and the data on other road segments, which essentially indicates the spatiotemporal correlation among the traffic status data.…”

Section: ) Lstm Based Secure Report Aggregationmentioning

confidence: 99%

TrafficChain: A Blockchain-Based Secure and Privacy-Preserving Traffic Map

Wang

Guo

et al. 2020

IEEE Access

Self Cite

View full text Add to dashboard Cite

Intelligent Connected Vehicles (ICVs) can provide smart, safe, and efficient transportation services and have attracted intensive attention recently. Obtaining timely and accurate traffic information is one of the most important problems in transportation systems, which would allow people to select fast routes and avoid congestions, thus saving their travel time on the road. Currently, the most popular ways to obtain traffic information is to inquire navigation agents, e.g., Apple map, and Google map. However, these navigation agents are essentially centralized systems, which are vulnerable to service congestions, a single point of failure, and attacks. Furthermore, users' privacy gets compromised as the agents can know their home and work addresses and hence their identities, track them in real-time, etc. In this paper, we propose TrafficChain, a secure and privacy-preserving decentralized traffic information collection system on the blockchain, by taking advantage of fog/edge computing infrastructure. In particular, we employ a two-layer blockchain architecture in TrafficChain to improve system efficiency, design a privacy-preserving scheme to protect users' identities and travel traces, and devise LSTM based deep learning mechanisms that can defend against Byzantine attacks and Sybil attacks in our system. Furthermore, an incentive mechanism is designed to motivate users to participate in the system. Simulation results show that TrafficChain works very efficiently and is resilient to both Byzantine attacks and Sybil attacks. INDEX TERMS Blockchain, intelligent connected vehicles (ICVs), fog/edge computing, Byzantine attacks, Sybil attacks, LSTM.

show abstract

“…proposed a model based on Word2vec word vector synthesis, which can be used for document synthesis simply based on word vectors. Peng H et al [10] and Ji J et al [11] proposed to use BiLSTM and CNN models for text representation, and to build more complex network models for document modeling, so as to obtain key information from documents.…”

Section: Introductionmentioning

confidence: 99%