An LSTM-Based Topic Flow Pattern Learning Algorithm and Its Application in Deceptive Review Detection
| Autor: | Shu-Juan Ji, Lu-yu Dong, Na Liu |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Artificial neural network
Relation (database) Computer science 02 engineering and technology 010501 environmental sciences Flow pattern 01 natural sciences Focus (linguistics) Discriminative model 0202 electrical engineering electronic engineering information engineering Benchmark (computing) 020201 artificial intelligence & image processing Construct (philosophy) Algorithm 0105 earth and related environmental sciences Meaning (linguistics) |
| Zdroj: | Advances in Intelligent Systems and Computing ISBN: 9789811358401 ICGEC |
| DOI: | 10.1007/978-981-13-5841-8_58 |
| Popis: | Great progress has been made in detecting deceptive reviews based on traditional machine learning. However, existing classification methods for detecting deceptive reviews mainly focus on combining word-embedding with the traditional machine learning algorithm, while neglecting the latent semantic meaning and its temporal relation of the topic in a review. In this paper, we propose a deceptive review detection model based on the learning of topic flow pattern. First, this paper analyzes the topic of reviews and considers the temporal relations of topics simultaneously to construct discriminative characteristics of reviews. Second, the Long Short-Term Memory (LSTM) neural network is used to classify reviews that contain information on topic series. Experimental results on three domains of datasets show that our proposed method is superior to benchmark methods. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|