A New Semantic Relations-Based Hybrid Approach for Implicit Aspect Identification in Sentiment Analysis

Abstract

Within the next few years, sentiment analysis or opinion mining is set to become an important component of real-world applications for product manufacturers, e-commerce companies, and potential customers. Sentiment analysis deals with the computational assessment of people’s opinions apparent or hidden within the text according to three levels: document, sentence and aspect levels. The aspect-level is increasingly becoming an active phase of sentiment analysis. At this level, the aim is to determine the hidden target of opinion represented in datasets, known as aspect term identification. This paper proposes an original hybrid model combining semantic relations and frequency-based approach with supervised classifiers for implicit aspect identification (IAI). The proposed approach is directed towards improving the F1-performances for traditional supervised classifiers commonly used in this field based on eager and lazy learning, and deep learning technique using long short-term memory whit attention mechanism applied for IAI. Particularly, this work addresses aspect term extraction and aggregation, the two sub-tasks of IAI, involving adjectives and verbs. The effects of this approach are empirically examined on multiple datasets of electronic products and restaurant reviews with multiple aspect granularity levels. Comparing this method with similar approaches clearly shows the benefits of this method: (i) the use of an appropriately selected WordNet semantic relations of adjectives and verbs that significantly helps classifiers for IAI. (ii) Using the hybrid model helps classifiers better handle these selected WordNet semantic relations and therefore deal better with IAI.

An Enhanced SVM Model for Implicit Aspect Identification in Sentiment Analysis

“…• In the case of Restaurant dataset, despite the difficulty of adjusting the core model that is more challenging and sensitive, our proposed technique (with the three kernels) shows a highly competitive performance level when compared to all works even the ones operating on data level which is less challenging and even the deep learning methods of [19] that are generally reputed for high classification performance.…”

“…The work in [17], is an earlier dictionary-based method to identify aspects conveyed by adjectives. The authors of [18] perform an implicit aspect identification task for adjectives and verbs using definition and synonym relations extracted from WordNet. In [19], authors propose a new hybrid model for implicit aspect identification that uses semantic relations combined with a frequency-based method and supervised classifiers.…”

“…They do, however, operate at various levels. While the techniques proposed in [18] and [19] concentrate on improving training data quality by acting at the data level which is a less challenging level, the approach proposed in [26] and our suggested method operate at the algorithmic level by suggesting modifications or additions. The works in [21], [17], [20] and [28] are hybrid methods that operate at both data level and algorithmic level.…”

Multi-Level Knowledge Engineering Approach for Mapping Implicit Aspects to Explicit Aspects

“…Similarly, El Hannach et al [32] also proposed a hybrid approach to identify implicit aspects from product reviews. Their approach first identifies candidates of implicit aspects and then maps them to any of the explicit aspects.…”

A Multi-Scale Local Attention Mechanism for Aspect Extraction

“…For instance, Avinash et al [8] used a hierarchical self-attention network to capture the importance of words and the internal dependencies within sentences, assisting in the recognition of attribute terms. Similarly, Hannach et al [9] applied LSTM with an attention mechanism for implicit aspect identification. However, current aspect extraction methods leveraging attention mechanisms exhibit two primary limitations.…”