Table 1. Analysis of past studies of feature extraction using pattern rules.
| Studies | Feature extraction methods | Datasets | Domain | Results |
|---|---|---|---|---|
| Pak & Günal (2022) | Sequential pattern-based rule mining | Hu & Liu (2004); SemEval 2014 (Task 4) | Electronic Products/Restaurant | F1: 70.0% |
| Rana et al. (2021) | Syntactic rules with opinion lexicon for Urdu language | Mukhtar et al. (2018) | Urdu opinion texts | P:78.0%, R: 76.0%, F1: 76.0% |
| Tran, Duangsuwan & Wettayaprasit (2021) | Aspect knowledge-based generation using pattern rules (AKGPR) | Hu & Liu (2004); Liu et al. (2016) | Electronic/ Computer products |
P:89.0%, R: 76.0%, F1: 81.0% P: 85.0%, R: 64.0%, F1: 73.0% |
| Tubishat, Idris & Abushariah (2021) | Heuristic Patterns, Whale Optimization Algorithm and Pruning | Hu & Liu (2004); Liu et al. (2015) | Electronic/computer products | P: 92.0%, R: 93.0%, F1: 92.0% |
| Chauhan & Meena (2020) | Domain-Specific aspect term extraction | Hu & Liu (2004) | Electronic Products | P: 88.0%, R: 85.0%, F1: 86.0% |
| Rana & Cheah (2019) | Sequential Pattern Rules using PrefixSpan algorithm with SPMF | Hu & Liu (2004) | Electronic Products | P: 86.0%, R: 91.0%, F1: 89.0% |
| Kang & Zhou (2017) | Extended DP with additional new rules | Hu & Liu (2004) | Electronic Products | P: 87.0%, R: 88.0%, F1: 87.0% |
| Liu et al. (2016) | Extended DP with Simulating Annealing | Hu & Liu (2004); Liu et al. (2015) | Electronic/ computer products | P: 85.0%, R: 91.0%, F1: 88.0% |
| Qiu et al. (2011) | Double Propagation (DP) using dependency rules and pruning | Hu & Liu (2004) | Electronic Products | P: 88.0%, R: 83.0%, F1: 86.0% |
| Khan et al. (2019) | EnSWF: POS and ngram-based ensemble method | Pang, Lee & Vaithyanathan (2002); Blitzer, Dredze & Pereira (2007); McAuley & Leskovec (2013) | Movie, Book, DVD, Electronics and Kitchen | Accuracy 91.64% |
| Asghar et al. (2019) | Heuristic patterns and lexicons | Hu & Liu (2004) | Electronic Products | P: 83.0%, R: 71.0%, F1: 77.0% |
| Agerri & Rigau (2019) | OTE using sequence labelling |
SemEval 2014 SemEval 2015 SemEval 2016 |
Customer Reviews | (SemEval 2014): P: 81.5%, R: 87.3%, F1: 84.1% (SemEval 2015): P: 72.9%, R: 69.0%, F1: 70.9% (SemEval 2016): P: 73.3%, R: 73.7%, F1: 73.5% |
| Konjengbam et al. (2018) | Aspect Ontology | Hu & Liu (2004) | Electronic Products | P: 79.0%, R: 79.0%, F1: 79.0% |
| Rana & Cheah (2017) | Two-fold-rule based method | Hu & Liu (2004) | Electronic Products | P: 87.0%, R: 92.0%, F1: 89.0% |
| Akhtar et al. (2017) | PSO based ensemble learning method | SemEval 2014 | Customer Reviews | P: 87.1%, R: 82.1%, F1: 84.5% |
| He et al. (2017) | Word embedding models with attention mechanism | Citysearch corpus | Restaurant Reviews | P: 85.7%, R: 72.2%, F1: 77.5% |
| Samha & Li (2016) | Dependency relations and lexicon | Hu & Liu (2004) | Electronic Products | P: 83.0%, R: 87.0%, F1: 77.0% |
| Khan & Jeong (2016) | Lazy learning model using syntactic rules | Hu & Liu (2004) | Electronic Products | P: 81.0%, R: 82.0% |
| Maharani, Widyantoro & Khodra (2015) | Pattern based extraction with new set of rules for explicit features | Hu & Liu (2004); Ding, Liu & Yu (2008) | Electronic Products | P: 62.6.0%, R: 72.8.0%, F1: 67.2% |
| Khan, Baharudin & Khan (2014) | Combined Pattern Based Noun Phrases (cBNP) | Hu & Liu (2004); Ferreira, Jakob & Gurevych (2008) | Electronic Products | P: 79.0%, R: 72.0%: F1:75.2% |
| Htay & Lynn (2013) | Pattern based extraction with new set of rules | Hu & Liu (2004) | Electronic Products | P: 73.0%, R: 86.0%, F1:79.0% |
| Ravi Kumar & Raghuveer (2013) | Dependencies using LingPipe Sentence Boundary, Lexicon and GI | amazon.com cnet.com |
Customer Reviews | P: 73.0%, R: 82.0% |
| Bagheri, Saraee & de Jong (2013) | Iterative bootstrapping using rules and pruning | Hu & Liu (2004) | Electronic Products | P: 86.0%, R: 64.0%, F1:73.0% |