Aspect-Category-Opinion-Sentiment Quadruple Extraction with Implicit Aspects and Opinions

Abstract

Product reviews contain a large number of implicit aspects and opinions. However, most of the existing studies in aspect-based sentiment analysis ignored this problem. In this work, we introduce a new task, named Aspect-Category-Opinion-Sentiment (ACOS) Quadruple Extraction, with the goal to extract all aspect-category-opinion-sentiment quadruples in a review sentence and provide full support for aspect-based sentiment analysis with implicit aspects and opinions. We further construct two new datasets Restaurant-ACOS and Laptop-ACOS for this new task. The former is an extension of the SemEval Restaurant dataset; the latter is a brand new Laptop dataset with much larger size than the Se-mEval Laptop dataset. Both contain the annotations of not only aspect-category-opinionsentiment quadruples but also implicit aspects and opinions. We finally benchmark the task with four baseline systems. Experiments demonstrate the feasibility of the new task and its advantage in extracting and describing implicit aspects and implicit opinions in ABSA. The two datasets and source code of four systems are publicly released at https:

A Survey on Aspect-Based Sentiment Analysis: Tasks, Methods, and Challenges

“…TwoStage [12], JET [91], GTS [75], OTE-MTL [92], BMRC [93] Dual-MRC [94], GAS [95], Gen-ABSA [96], Span-ASTE [97], NAG-ASTE [98], PASTE [99] Aspect-Category-Sentiment Detection (ACSD) TAS-BERT [46], MEJD [100], GAS [95], Paraphrase [7] Quad Extraction Aspect Sentiment Quad Prediction (ASQP) Extract-Classify-ACOS [101], Paraphrase [7] Fig. 2.…”

“…where the encoder Enc(•) encodes contextualized features of the input, the decoder Dec(•) generates a token at each step, based on the encoded input and the previous output. [111] English laptops, restaurants https://alt.qcri.org/semeval2015/task12/ SemEval-2016 [112] multilingual electronics, hotels, restaurants https://alt.qcri.org/semeval2016/task5/ TOWE [45] English laptops, restaurants https://github.com/NJUNLP/TOWE ASC-QA [113] Chinese bags, cosmetics, electronics https://github.com/jjwangnlp/ASC-QA MAMS [114] English restaurants https://github.com/siat-nlp/MAMS-for-ABSA ARTS [115] English laptops, restaurants https://github.com/zhijing-jin/ARTS TestSet ASTE-Data-V2 [91] English laptops, restaurants https://github.com/xuuuluuu/Position-Aware-Tagging-for-ASTE ASAP [116] Chinese restaurants https://github.com/Meituan-Dianping/asap ACOS [101] English laptops, restaurants https://github.com/NUSTM/ACOS ABSA-QUAD [7] English restaurants https://github.com/IsakZhang/ABSA-QUAD…”

“…With these annotations available, Xu et al [91] organize them together (with minor refinements) to obtain the ASTE-Data-V2 dataset, which contains (aspect term, opinion term, sentiment polarity) triplets for each sample sentence 4 . Most recently, two datasets including ACOS [101] and ABSA-QUAD [7] for the sentiment quad prediction task have been introduced, where each data instance of them contains the annotations of the four sentiment elements in the quad format.…”

“…Although they can be useful under different scenarios, a model which can predict the four sentiment elements in one shot is supposed to provide the most complete aspect-level sentiment structure. This leads to the aspect sentiment quad prediction (ASQP) task proposed recently 8 [101,7], aiming to predict all the four sentiment elements in the quadruplet form given a text item. Returning to the example in Table 2, two sentiment quads are expected: (food, pizza, POS, delicious) and (service, service, NEG, terrible).…”

“…Cai et al [101] study the ASQP task with an emphasis on the implicit aspects or opinions. The authors argue that implicit aspects or opinions appear frequently in real-world scenarios, and use "null" to denote them in the sentiment quads.…”