{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,3]],"date-time":"2026-01-03T15:21:51Z","timestamp":1767453711920,"version":"3.41.2"},"reference-count":50,"publisher":"Emerald","issue":"1","license":[{"start":{"date-parts":[[2022,5,9]],"date-time":"2022-05-09T00:00:00Z","timestamp":1652054400000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.emerald.com\/insight\/site-policies"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["AJIM"],"published-print":{"date-parts":[[2023,1,6]]},"abstract":"<jats:sec><jats:title content-type=\"abstract-subheading\">Purpose<\/jats:title><jats:p>To understand the meaning of a sentence, humans can focus on important words in the sentence, which reflects our eyes staying on each word in different gaze time or times. Thus, some studies utilize eye-tracking values to optimize the attention mechanism in deep learning models. But these studies lack to explain the rationality of this approach. Whether the attention mechanism possesses this feature of human reading needs to be explored.<\/jats:p><\/jats:sec><jats:sec><jats:title content-type=\"abstract-subheading\">Design\/methodology\/approach<\/jats:title><jats:p>The authors conducted experiments on a sentiment classification task. Firstly, they obtained eye-tracking values from two open-source eye-tracking corpora to describe the feature of human reading. Then, the machine attention values of each sentence were learned from a sentiment classification model. Finally, a comparison was conducted to analyze machine attention values and eye-tracking values.<\/jats:p><\/jats:sec><jats:sec><jats:title content-type=\"abstract-subheading\">Findings<\/jats:title><jats:p>Through experiments, the authors found the attention mechanism can focus on important words, such as adjectives, adverbs and sentiment words, which are valuable for judging the sentiment of sentences on the sentiment classification task. It possesses the feature of human reading, focusing on important words in sentences when reading. Due to the insufficient learning of the attention mechanism, some words are wrongly focused. The eye-tracking values can help the attention mechanism correct this error and improve the model performance.<\/jats:p><\/jats:sec><jats:sec><jats:title content-type=\"abstract-subheading\">Originality\/value<\/jats:title><jats:p>Our research not only provides a reasonable explanation for the study of using eye-tracking values to optimize the attention mechanism but also provides new inspiration for the interpretability of attention mechanism.<\/jats:p><\/jats:sec>","DOI":"10.1108\/ajim-12-2021-0385","type":"journal-article","created":{"date-parts":[[2022,5,6]],"date-time":"2022-05-06T15:00:06Z","timestamp":1651849206000},"page":"20-43","source":"Crossref","is-referenced-by-count":1,"title":["Does attention mechanism possess the feature of human reading? A\u00a0perspective of sentiment classification task"],"prefix":"10.1108","volume":"75","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-9077-5720","authenticated-orcid":false,"given":"Lei","family":"Zhao","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yingyi","family":"Zhang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9522-2914","authenticated-orcid":false,"given":"Chengzhi","family":"Zhang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"140","published-online":{"date-parts":[[2022,5,9]]},"reference":[{"first-page":"1","article-title":"Eyes on the parse: using gaze features in syntactic parsing","year":"2020","key":"key2023010703144631000_ref001"},{"year":"2015","key":"key2023010703144631000_ref002","article-title":"Neural machine translation by jointly learning to align and translate"},{"first-page":"302","article-title":"Sequence classification with human attention","year":"2018","key":"key2023010703144631000_ref003"},{"first-page":"579","article-title":"Weakly supervised part-of-speech tagging using eye-tracking data","year":"2016","key":"key2023010703144631000_ref004"},{"issue":"11","key":"key2023010703144631000_ref005","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1111\/lnc3.12396","article-title":"Sequence labelling and sequence classification with gaze: novel uses of eye-tracking data for natural language processing","volume":"14","year":"2020","journal-title":"Language and Linguistics Compass"},{"first-page":"1442","article-title":"Massive exploration of neural machine translation architectures","year":"2017","key":"key2023010703144631000_ref006"},{"year":"2019","key":"key2023010703144631000_ref007","article-title":"What does BERT look at? An analysis of BERT's attention\u201d, ArXiv Preprint ArXiv:1906.04341"},{"first-page":"4171","article-title":"BERT: pre-training of deep bidirectional transformers for language understanding","year":"2019","key":"key2023010703144631000_ref008"},{"key":"key2023010703144631000_ref009","unstructured":"Goldberg, Y. (2019), \u201cAssessing BERT's syntactic abilities\u201d, ArXiv Preprint ArXiv:1901.05287, available at: http:\/\/arxiv.org\/abs\/1901.05287."},{"first-page":"2047","article-title":"Framewise phoneme classification with bidirectional LSTM networks","year":"2005","key":"key2023010703144631000_ref010"},{"issue":"6","key":"key2023010703144631000_ref011","doi-asserted-by":"crossref","first-page":"739","DOI":"10.1108\/AJIM-02-2019-0050","article-title":"Using the eye-tracking method to study consumer online health information search behaviour","volume":"71","year":"2019","journal-title":"Aslib Journal of Information Management"},{"issue":"8","key":"key2023010703144631000_ref012","doi-asserted-by":"crossref","first-page":"1735","DOI":"10.1162\/neco.1997.9.8.1735","article-title":"Long short-term memory","volume":"9","year":"1997","journal-title":"Neural Computation"},{"first-page":"1","article-title":"Entity recognition at first sight: improving NER with eye movement information","year":"2019","key":"key2023010703144631000_ref013"},{"issue":"1","key":"key2023010703144631000_ref014","first-page":"1","article-title":"Data descriptor: ZuCo, a simultaneous EEG and eye-tracking resource for natural sentence reading","volume":"5","year":"2018","journal-title":"Scientific Data"},{"year":"2020","key":"key2023010703144631000_ref015","article-title":"exBERT: a visual analysis tool to explore learned representations in transformer models\u201d, ArXiv Preprint ArXiv:1910.05276"},{"key":"key2023010703144631000_ref016","unstructured":"Htut, P.M., Phang, J., Bordia, S. and Bowman, S.R. (2019), \u201cDo attention heads in BERT track syntactic dependencies?\u201d, ArXiv Preprint ArXiv:1911.12246, available at: http:\/\/arxiv.org\/abs\/1911.12246."},{"year":"2019","key":"key2023010703144631000_ref017","article-title":"Attention is not explanation"},{"first-page":"427","article-title":"Bag of tricks for efficient text classification","year":"2017","key":"key2023010703144631000_ref018"},{"issue":"1","key":"key2023010703144631000_ref019","doi-asserted-by":"crossref","first-page":"79","DOI":"10.1214\/aoms\/1177729694","article-title":"On information and sufficiency","volume":"22","year":"1951","journal-title":"The Annals of Mathematical Statistics"},{"first-page":"4791","article-title":"Has machine translation achieved human parity? A case for document-level evaluation","year":"2020","key":"key2023010703144631000_ref020"},{"volume-title":"Interpretable Neural Models for Natural Language Processing","year":"2017","key":"key2023010703144631000_ref021"},{"issue":"4","key":"key2023010703144631000_ref022","doi-asserted-by":"crossref","first-page":"459","DOI":"10.1007\/s12559-019-9624-y","article-title":"Improving user attribute classification with text and social network attention","volume":"11","year":"2019","journal-title":"Cognitive Computation"},{"issue":"1","key":"key2023010703144631000_ref023","doi-asserted-by":"crossref","first-page":"145","DOI":"10.1109\/18.61115","article-title":"Divergence measures based on the Shannon entropy","volume":"37","year":"1991","journal-title":"IEEE Transactions on Information Theory"},{"key":"key2023010703144631000_ref024","doi-asserted-by":"crossref","first-page":"325","DOI":"10.1016\/j.neucom.2019.01.078","article-title":"Bidirectional LSTM with attention mechanism and convolutional layer for text classification","volume":"337","year":"2019","journal-title":"Neurocomputing"},{"first-page":"462","article-title":"A cognition based attention model for sentiment analysis","year":"2017","key":"key2023010703144631000_ref025"},{"year":"2020","key":"key2023010703144631000_ref026","article-title":"Bridging information-seeking human gaze and machine reading comprehension\u201d, ArXiv Preprint ArXiv:2009.14780"},{"first-page":"81","article-title":"How human analyse lexical indicators of sentiments \u2013 a cognitive analysis using reaction-time","year":"2012","key":"key2023010703144631000_ref027"},{"volume-title":"Distribution-Free Statistical Methods","year":"1995","key":"key2023010703144631000_ref028"},{"first-page":"3747","article-title":"Predicting readers' sarcasm understandability by modeling gaze behavior","year":"2016","key":"key2023010703144631000_ref029"},{"first-page":"156","article-title":"Leveraging cognitive features for sentiment analysis","year":"2016","key":"key2023010703144631000_ref030"},{"first-page":"5884","article-title":"Cognition-cognizant sentiment analysis with multitask subjectivity summarization based on annotators' gaze behavior","year":"2018","key":"key2023010703144631000_ref031"},{"first-page":"1101","article-title":"Explainable prediction of medical codes from clinical text","year":"2018","key":"key2023010703144631000_ref032"},{"first-page":"1532","article-title":"GloVe: global vectors for word representation","year":"2014","key":"key2023010703144631000_ref033"},{"key":"key2023010703144631000_ref034","doi-asserted-by":"publisher","DOI":"10.1093\/acprof:oso\/9780198566816.003.0006","article-title":"Eye movement control in reading and the E-Z reader model","volume-title":"Cognitive Processes in Eye Guidance","year":"2012"},{"first-page":"601","article-title":"Using gaze data to predict multiword expressions","year":"2017","key":"key2023010703144631000_ref035"},{"key":"key2023010703144631000_ref036","unstructured":"Rong, Y., Xu, W., Akata, Z. and Kasneci, E. (2021), \u201cHuman attention in fine-grained classification\u201d, ArXiv Preprint ArXiv:2111.01628, available at: http:\/\/arxiv.org\/abs\/2111.01628."},{"issue":"3","key":"key2023010703144631000_ref037","doi-asserted-by":"crossref","first-page":"325","DOI":"10.1108\/AJIM-07-2018-0182","article-title":"The influence of snippet length on user behavior in mobile web search: an experimental eye-tracking study","volume":"71","year":"2019","journal-title":"Aslib Journal of Information Management"},{"first-page":"4596","article-title":"Human attention maps for text classification: do humans and neural networks focus on the same words?","year":"2020","key":"key2023010703144631000_ref038"},{"issue":"7","key":"key2023010703144631000_ref039","doi-asserted-by":"crossref","first-page":"7363","DOI":"10.1007\/s12652-020-02412-1","article-title":"An abstractive summary generation system for customer reviews and news article using deep learning","volume":"12","year":"2021","journal-title":"Journal of Ambient Intelligence and Humanized Computing"},{"first-page":"1631","article-title":"Recursive deep models for semantic compositionality over a sentiment treebank","year":"2013","key":"key2023010703144631000_ref040"},{"article-title":"Improving natural language processing tasks with human gaze-guided neural attention","volume-title":"Advances in Neural Information Processing Systems","year":"2020","key":"key2023010703144631000_ref041"},{"issue":"6","key":"key2023010703144631000_ref042","first-page":"937","article-title":"A note on one class of perceptrons","volume":"25","year":"1964","journal-title":"Automation and Remote Control"},{"key":"key2023010703144631000_ref044","first-page":"5998","article-title":"Attention is all you need","volume":"30","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"year":"2018","key":"key2023010703144631000_ref045","article-title":"Attention-based transactional context embedding for next-item recommendation"},{"article-title":"Attention is not explanation","year":"2020","key":"key2023010703144631000_ref046","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/d19-1002"},{"issue":"4","key":"key2023010703144631000_ref047","doi-asserted-by":"crossref","first-page":"434","DOI":"10.1108\/AJIM-01-2018-0017","article-title":"Emotions and eye-tracking of differing age groups searching on e-book wall","volume":"70","year":"2018","journal-title":"Aslib Journal of Information Management"},{"issue":"2","key":"key2023010703144631000_ref048","doi-asserted-by":"crossref","first-page":"703","DOI":"10.1007\/s11192-021-04230-4","article-title":"Enhancing keyphrase extraction from academic articles with their reference information","volume":"127","year":"2022","journal-title":"Scientometrics"},{"first-page":"957","article-title":"Human-like explanation for text classification with limited attention supervision","year":"2022","key":"key2023010703144631000_ref049"},{"issue":"5","key":"key2023010703144631000_ref050","doi-asserted-by":"crossref","first-page":"611","DOI":"10.1002\/asi.24430","article-title":"Enhancing keyphrase extraction from microblogs using human reading time","volume":"72","year":"2021","journal-title":"Journal of the Association for Information Science and Technology"},{"first-page":"425","article-title":"Human behavior inspired machine reading comprehension","year":"2019","key":"key2023010703144631000_ref051"}],"container-title":["Aslib Journal of Information Management"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.emerald.com\/insight\/content\/doi\/10.1108\/AJIM-12-2021-0385\/full\/xml","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.emerald.com\/insight\/content\/doi\/10.1108\/AJIM-12-2021-0385\/full\/html","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,7,24]],"date-time":"2025-07-24T23:01:35Z","timestamp":1753398095000},"score":1,"resource":{"primary":{"URL":"http:\/\/www.emerald.com\/ajim\/article\/75\/1\/20-43\/60683"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2022,5,9]]},"references-count":50,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2022,5,9]]},"published-print":{"date-parts":[[2023,1,6]]}},"alternative-id":["10.1108\/AJIM-12-2021-0385"],"URL":"https:\/\/doi.org\/10.1108\/ajim-12-2021-0385","relation":{},"ISSN":["2050-3806"],"issn-type":[{"type":"print","value":"2050-3806"}],"subject":[],"published":{"date-parts":[[2022,5,9]]}}}