Call for Papers 2024 |
Feb 2024 - Volume 15, Issue 1
Deadline: 15 Jan 2024
Publication: 20 Feb 2024
Apr 2024 - Volume 15, Issue 2
Deadline: 15 Mar 2024
Publication: 20 Apr 2024
More
|
|
|
ABSTRACT
Title |
: |
A HYBRID DEEP NEURAL NETWORK FOR ASPECT BASED SENTIMENT ANALYSIS ON RAJYA SABHA QUESTIONS |
Authors |
: |
Shreyas R Hegde, Yogesh R Gaikwad |
Keywords |
: |
Aspect Based Sentiment Analysis; Attention-Based Aspect Extraction; Long Short-Term Memory; Target Dependent; Target Control; Attention Term Aspect Extraction. |
Issue Date |
: |
Jan-Feb 2021 |
Abstract |
: |
In the modern era, Technology and Politics are strongly connected and have become inseparable. This paper proposes a hybrid approach by using unsupervised and supervised techniques for analyzing the innumerable opinions expressed in the Rajya Sabha Question Hour to extract aspects, sentiments and perform Aspect Based Sentiment Analysis (ABSA). An unsupervised Attention-Based Aspect Extraction-Long Short-Term Memory (ABAE-LSTM) network is used to identify the āNā cluster of aspects present in the corpus and categorize the aspect terms present in the questions into one of these aspect categories. Supervised Neural Networks, Target Dependent-LSTM (TD-LSTM), Target Control-LSTM (TC-LSTM) and Attention Term Aspect Extraction-LSTM (ATAE-LSTM) are employed to perform ABSA. Experiments conducted on the unsupervised ABAE-LSTM show high coherence scores between the aspect terms, and this also creates a gold standard training data set with aspect and sentiment labels in the domain. Results from supervised techniques then display promising accuracy for all three models. |
Page(s) |
: |
112-128 |
ISSN |
: |
0976-5166 |
Source |
: |
Vol. 12, No.1 |
PDF |
: |
Download |
DOI |
: |
10.21817/indjcse/2021/v12i1/211201199 |
|