CoChat
Sign Up
Cited by 622Sentiment Analysis and Opinion MiningDiverse Aspects of Tourism ResearchDigital Marketing and Social Media

Sentiment Analysis in Tourism: Capitalizing on Big Data

Open access
Alaei, Ali Reza, Becken, Susanne, Stantic, Bela·2019·Journal of Travel Research
DOI 10.1177/0047287517747753
Open this paper in CoChat — free accountOpen PDF

Abstract

Advances in technology have fundamentally changed how information is produced and consumed by all actors involved in tourism. Tourists can now access different sources of information, and they can generate their own content and share their views and experiences. Tourism content shared through social media has become a very influential information source that impacts tourism in terms of both reputation and performance. However, the volume of data on the Internet has reached a level that makes manual processing almost impossible, demanding new analytical approaches. Sentiment analysis is rapidly emerging as an automated process of examining semantic relationships and meaning in reviews. In this article, different sentiment analysis approaches applied in tourism are reviewed and assessed in terms of the datasets used and performances on key evaluation metrics. The article concludes by outlining future research avenues to further advance sentiment analysis in tourism as part of a broader Big Data approach.

AI snapshot

Written from the paper’s full text.

TL;DR

This review examines sentiment analysis methods used in tourism, assesses their datasets and performance, and argues that sentiment analysis is an early step toward broader Big Data approaches in tourism research. It finds that machine learning and dictionary-based methods dominate, hotel/accommodation reviews are the most common data source, and multi-class or neutral sentiment detection remains difficult.

Key findings

  • Tourism sentiment studies rely heavily on hotel and accommodation reviews, with fewer studies on restaurants and airlines; common sources include TripAdvisor, Booking, Ctrip, and Twitter. Reported dataset sizes vary widely, from 479 Chinese hotel reviews in Zheng and Ye to 417,170 hotel reviews in Brob, and 10,000,000 tweets in Shimada et al. with 200,000 or 116 annotated items depending on the task.
  • Binary classification generally performs better than three-class classification. Examples include Zheng and Ye’s SVM on hotel reviews at 0.912 accuracy, Zhang et al.’s Naïve Bayes on restaurant reviews at 0.957 accuracy, and Shimada et al.’s Naïve Bayes on tourism information at 0.92 accuracy; three-class results are lower, such as Kasper and Vela at 0.81 accuracy, Schmunk et al. at 0.768, and Bucur at 0.72.
  • On the Sanders Twitter dataset, Levallois’s method gave the best binary results (accuracy 0.82, F-measure 0.82), while VADER gave the best multi-class results (accuracy 0.60, F-measure 0.58).
  • The review reports that machine learning approaches are most common, dictionary-based approaches are also frequent, and hybrid approaches are rare; domain-specific lexicons are seldom used, and neutral reviews are especially hard to detect.

Methods

  • The paper is a critical review of sentiment analysis in tourism. Studies were identified through Google keyword searches and by extracting tourism-related references from recent sentiment analysis reviews, rather than through a systematic Scopus or Web of Science search.
  • It provides a technical overview of sentiment analysis steps—data retrieval, extraction and selection, pre-processing, feature extraction, topic detection, and data mining—and categorizes methods as supervised machine learning, unsupervised machine learning, dictionary/rule-based, semantic, and hybrid.
  • It compares reported tourism studies using accuracy, precision, recall, and F-measure, and additionally evaluates three Twitter-oriented methods on a publicly available Twitter dataset.

Limitations

  • Stated in the paper: results across tourism studies are not directly comparable because database sizes, numbers of classes, and data types differ; the search strategy was not a systematic database search; domain-specific lexicons are rarely used, which compromises quality; neutral reviews are difficult to detect; and hybrid methods have been used only in a few cases.
  • Stated in the paper: some dictionary-based tourism studies were restricted by small keyword sets or failure to handle negation, limiting their findings.
  • Evident from the text: many studies in the summary tables do not report key dataset or performance details, and the review does not conduct a meta-analysis or standardized re-evaluation of all included tourism methods.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1177/0047287517747753

Cites

89 of 89 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

14 of 144 cited sources checked · 8 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    Social media and blogs nowadays complement opinions attained from relatives, friends, colleagues, and official sources.

    “Therefore, social media and blogs nowadays complements opinions attained from relatives, friends, colleagues, and official sources (Cantallops & Salvi, 2014;Chua & Banerjee, 2013;Filieri et al., 2015;Hepburn, 2007;Mack et al., 2008).”
    Quote in the paperNot a substantial claim1 of 3 sources checked
    • cites Cantallops & Salvi, 2014 source
    • cites Chua & Banerjee, 2013

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Filieri et al., 2015 source

      Section 2 says travelers use online reviews and treats consumer-generated content as important alongside officially provided information. It does not establish that social media and blogs complement opinions from relatives, friends, and colleagues; section 3.4.2 instead discusses travelers sharing online-review tips with friends. (LLM, conf=0.79)

      Experimental — accuracy not yet measured
  • Interpretationcitation

    Integration of multiple Big Data sources may reveal new insights that were not able to be detected with traditional approaches.

    “Integration of multiple Big Data sources, e.g., heterogeneous data sources, in the form of structured and unstructured data, such as customer feedback, reservation and booking data, and web search / navigation data, in customer and supplier sides (Höpken et al., 2013, Höpken et al., 2015), may reveal new insights that were not able to be detected with traditional approaches.”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Höpken et al., 2013 source
    • cites Höpken et al., 2015 source
  • Premisecitation

    Sentiment analysis in tourism is only beginning to gain in popularity.

    “Whilst highly relevant for tourism, sentiment analysis in tourism is only beginning to gain in popularity (Feldman, 2013, Gao et al., 2015, Ribeiro et al., 2016).”
    Quote in the paperNot a substantial claim1 of 3 sources checked
    • cites Feldman, 2013 source
    • cites Gao et al., 2015 source
    • cites Ribeiro et al., 2016 source

      The Introduction discusses growing interest in sentiment analysis generally, and the Concluding remarks describe its use on social-network content. Neither addresses whether sentiment analysis in tourism was only beginning to gain popularity; searches for tourism, travel, and hotel applications found no relevant passage. Our copy therefore cannot verify the tourism-specific claim. (LLM, conf=0.77)

      Experimental — accuracy not yet measured
  • Findingown data

    The method proposed by Levallois (2013) provided the best results in the binary classification problem.

    “The method proposed by Levallois, (2013) has provided the best results in the binary classification problem.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Levallois, 2013 source

      The Accuracy section reports results for a three-class task—positive, negative, or neutral—and says another participant performed better on that task. Our copy does not report a binary-classification comparison, so it cannot establish whether Levallois’s method produced the best binary results. (LLM, conf=0.75)

      Experimental — accuracy not yet measured
  • Findingown data

    VADER (Hutto and Gilbert, 2014) delivered the best results in the multi-class classification case.

    “However, the VADER (Hutto and Gilbert, 2014) delivered the best results in the multi-class classification case.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Hutto and Gilbert, 2014 source

      The Introduction and Results report strong three-class classification performance on tweets: VADER achieved F1 = 0.96, above individual human raters’ F1 = 0.84. But the Results say VADER performed “as well or better” than machine-learning approaches in three of four domain comparisons, not that it delivered the best multiclass results across the board. No page numbers are available in our copy. (LLM, conf=0.83)

      Experimental — accuracy not yet measured
  • Classified as probably not a substantial claim. 73
  • May combine two claims. 26

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Umigon is a dictionary-based method that uses a lexicon with heuristics for sentiment detection in Twitter reviews.

    “Umigon is another dictionary -based method, which uses a lexicon with heuristics for sentiment detection in Twitter reviews (Levallois, 2013).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Levallois, 2013 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    In the data mining process, different types of sentiment analysis methods can be distinguished: machine learning, rule/dictionary-based and hybrid approaches.

    “In the data mining process, different types of sentiment analysis methods can be distinguished in the literature; namely (i) machine learning, (ii) rule/dictionary-based and (iii) hybrid approaches (Feldman, 2013;Ribeiro et al., 2016).”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites Feldman, 2013 source
    • cites Ribeiro et al., 2016 source

  • May combine two claims.

    Premisecitation

    Neural network and deep learning models and k-nearest neighbour method were also employed for semantic analysis in the literature.

    “Neural network and deep learning models (Irsoy & Cardie, 2014;Socher et al., 2013) and k-nearest neighbour method (Schmunk et al., 2014) were also employed for semantic analysis in the literature.”
    Quote in the paperNot a substantial claim2 of 3 sources checked
    • cites Irsoy & Cardie, 2014 source
    • cites Socher et al., 2013 source
    • cites Schmunk et al., 2014

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    VADER is a method that has provided promising results on Twitter data.

    “Valence Aware Dictionary for Sentiment Reasoning (VADER) is a method that has provided promising results on Twitter data (Hutto & Gilbert, 2014).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Hutto & Gilbert, 2014 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The VADER sentiment lexicon is composed of more than 7,000 items, along with their associated sentiment intensity measures, validated by humans and specifically adapted to sentiment in microblog-like contexts, such as Twitter.

    “The VADER sentiment lexicon is composed of more than 7,000 items, along with their associated sentiment intensity measures, validated by humans and specifically adapted to sentiment in microblog-like contexts, such as Twitter (Hutto & Gilbert, 2014).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Hutto & Gilbert, 2014 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Umigon provides additional semantic features, such as time or subjectivity.

    “Umigon provides additional semantic features, such as time or subjectivity (Levallois, 2013).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Levallois, 2013 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Some studies divided sentiment analysis into two subtasks: classifying sentences into objective and subjective, then determining polarity of subjective sentences.

    “Some studies followed a slightly different approach, whereby the sentiment analysis was divided into two subtasks, namely: (i) classifying sentences into objective and subjective sentences, and (ii) then determining the polarity of the subjective sentences (Marrese-Taylora et al., 2013;Riloff & Wiebe, 2003).”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites Marrese-Taylora et al., 2013

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Riloff & Wiebe, 2003 source

      Section 2.1 (page unavailable) says the paper’s goal is to classify individual sentences as subjective or objective. It does not describe a second step that determines the polarity of those subjective sentences; its discussion of positive and negative classification concerns other work on reviews at the document level. (LLM, conf=0.88)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S4 (based).

    Findingnone given

    None of the reviews to date focus on tourism.

    “None of the reviews to date focus on tourism.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Different types of features, such as term frequency, TF-IDF, stemmed word, bag-of-words, information gain, and n-gram, were proposed to characterise tourism reviews.

    “Different types of features, such as term frequency (TF) (Ye et al., 2009), term frequency-inverse document frequency (TF-IDF) (Shi & Li, 2011), stemmed word (Ganu et al., 2009), bag-ofwords (Markopoulos et al., 2015), information gain (IG) (Zheng & Ye, 2009), n-gram (Brob, 2013;Kang et al., 2012;Markopoulos et al., 2015;Zhang et al., 2011;Pablos et al., 2015) were proposed to characterise tourism”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Ye et al., 2009

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Shi & Li, 2011 source
    • cites Ganu et al., 2009

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Tourism researchers have typically used two types of online content for sentiment analysis: reviews from professional tourism websites (e.g., TripAdvisor, Booking, Ctrip) and social media posts (e.g., Twitter).

    “Tourism researchers have typically used two types of online content for their sentiment analysis: reviews of tourism obtained from professional websites (e.g., TripAdvisor, Booking, and Ctrip) and social media posts (e.g., Twitter).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    The Naïve Bayes sentiment classification approach in Shimada et al. was trained using automatically labelled data.

    “The Naïve Bayes sentiment classification approach was trained using automatically labelled data.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Shimada et al., 2011 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    Reviews containing a smiley face were considered positive and those with an angry face were classed as negative.

    “Therefore, reviews that contained a smiley face, for example, were considered as positive and those with an angry face were classed as negative (Shimada et al., 2011).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Shimada et al., 2011 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    A small number of tourism sentiment analysis studies focus on restaurants and airlines.

    “A small number of studies focus on restaurants (Ganu et al., 2009;Zhang et al., 2011) and airlines (Misopoulos et al., 2014).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Ganu et al., 2009

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Zhang et al., 2011 source
    • cites Misopoulos et al., 2014 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    An unsupervised machine learning approach based on Naïve Bayes classifier was implemented by Shimada et al. (2011) to produce a sentiment analysis of tourism data at the sentence level.

    “An unsupervised machine learning approach based on Naïve Bayes classifier was implemented by Shimada et al. (2011) to produce a sentiment analysis of tourism data at the sentence level.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Shimada et al., 2011 source

  • May combine two claims.

    Findingnone given

    Both supervised and unsupervised machine learning, dictionary based, semantic and hybrid sentiment analysis approaches were used in the tourism literature.

    “Both supervised and unsupervised machine learning, dictionary based, semantic and hybrid sentiment analysis approaches were used in the tourism literature.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingreasoning

    Both professional tourism review sites and social media posts usually contain short text.

    “Both types of sources usually contain short text.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Schmunk et al. (2014) discussed a system to initially detect the subjectivity of a sentence by a dictionary-based method, then classify the sentence into positive or negative using bigrams features along with a SVM classifier.

    “Schmunk et al. (2014) further discussed a system to initially detect the subjectivity of a sentence by a dictionary-based method.Then, the classification of the sentence into positive or negative was performed using bigrams features along with a SVM classifier (Schmunk et al., 2014).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Schmunk et al., 2014

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Waldhör and Rind (2008) proposed combining a linguistic parsing methodology with information and terminology extraction methods to determine sentiment polarity of online blog reviews.

    “Earlier work by Waldhör and Rind (2008) proposes to combine a linguistic parsing methodology with information and terminology extraction methods in order to determine sentiment polarity of online blog reviews.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Waldhör and Rind, 2008 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Kasper & Vela (2011 and 2012) presented a sentence level hybrid sentiment analysis system in the context of German language hotel reviews.

    “Another sentence level hybrid sentiment analysis system was presented by Kasper &Vela (2011 and2012) in the context of German language hotel reviews.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Kasper &Vela, 2011 and2012

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    The dictionary used in Gräbner et al.'s (2012) study was a hotel domain specific lexicon of semantically relevant words.

    “The dictionary used in Gräbner et al.'s (2012) study was a hotel domain specific lexicon of semantically relevant words.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Gräbner et al., 2012 source

  • May combine two claims.

    Findingcitation

    Most studies have used data written in English for sentiment analysis, but few used reviews written in Chinese, Spanish and German.

    “Most studies have used data written in English for sentiment analysis, but few used reviews written in Chinese, Spanish and German (Zheng & Ye, 2009;Tan & Wu, 2011;Garcia et al, 2012;Kasper & Vela, 2011;Zhang et al., 2011).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Zheng & Ye, 2009 source
    • cites Tan & Wu, 2011 source
    • cites Garcia et al, 2012

  • May combine two claims.

    Findingcitation

    Misopoulos et al.'s analysis was based on a limited number of 20 keywords (10 positive and 10 negative), which posed a significant restriction to the findings.

    “The analysis was, however, based on a limited number of 20 keywords (10 positive and 10 negative), which posed a significant restriction to the findings from this research.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Misopoulos et al., 2014 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    K-mean clustering techniques and statistical models based on probability distribution of reviews in sentiment space were employed on tourism data.

    “K-mean clustering techniques and statistical models based on probability distribution of reviews in sentiment space (Rossetti et al., 2015, Xiang et al., 2015b) were also employed on tourism data.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Rossetti et al., 2015 source
    • cites Xiang et al., 2015b

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Sharma et al. (2015) analysed travel reviews at the sentence level using a lexicon based system.

    “Another example can be found in Sharma et al. (2015) who analysed travel reviews at the sentence level using a lexicon based system.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Sharma et al., 2015

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Chiu et al. (2015) proposed sentiment analysis on Chinese e-WOM.

    “Sentiment analysis on Chinese e-WOM was proposed by Chiu et al. (2015).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Chiu et al., 2015

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    A relatively broad range of studies exist in the domain of tourism, mainly in relation to hotels and accommodation.

    “In summary, a relatively broad range of studies exist in the domain of tourism, mainly in relation to hotels and accommodation.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Our review revealed that most tourism sentiment analyses are based on a machine learning approach, although a considerable number of studies have also used a dictionary-based approach.

    “Furthermore, our review revealed that most tourism sentiment analyses are based on a machine learning approach, although a considerable number of studies have also used a dictionary-based approach.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Combining a supervised probabilistic model and a heuristic n-phrase rule was used to effectively obtain customer opinions about hotel attributes.

    “Combining a supervised probabilistic model and a heuristic n-phrase rule was used to effectively obtain customer opinions about hotel attributes.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Chiu et al., 2015

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Negation was not incorporated in Misopoulos et al.'s system to accurately capture the meaning of opinions such as not bad.

    “Moreover, negation was not incorporated in the system to accurately capture the meaning of opinions, such as”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Misopoulos et al., 2014 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Xiang et al. (2015a) proposed a semantic approach for text analysis to understand hotel guest experience and their satisfaction.

    “Xiang et al. (2015a) proposed semantic approach for text analysis to understand hotel guest experience and their satisfaction.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Xiang et al., 2015a

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Misopoulos et al.'s results revealed aspects of airline customer service where customers were dissatisfied, satisfied, or delighted.

    “The results revealed those aspects of the airline customer service where customers were dissatisfied, satisfied, or even delighted.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Misopoulos et al., 2014 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Other dictionary-based analysis focused on hotel and restaurant customer reviews.

    “Other dictionary-based analysis focused on hotel and restaurant customer reviews (Bucur, 2015;Gräbner et al., 2012;Marrese-Taylora et al., 2013;Schmunk et al., 2014).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Bucur, 2015 source
    • cites Gräbner et al., 2012 source
    • cites Marrese-Taylora et al., 2013

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Using binary choice keywords and a Naïve Bayes algorithm helped measure sentiment polarities of tweets related to different tourist destinations.

    “Using binary choice keywords and a Naïve Bayes algorithm helped measure sentiment polarities of tweets related to different tourist destinations (Claster et al., 2010a;Claster et al., 2010b;Claster et al., 2013).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Claster et al., 2010a

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Claster et al., 2010b

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Claster et al., 2013 source

  • May combine two claims.

    Findingreasoning

    The majority of tourism sentiment analyses used a machine learning approach, often trained with small annotated datasets.

    “Based on the above insights, it can be noted that the majority of tourism sentiment analyses used a machine learning approach, often trained with small annotated datasets, as this process needs considerable human resources.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Satisfaction as a theoretical construct has been explored and discussed for a long time, and multiple instruments exist to operationalise and measure it.

    “Satisfaction as a theoretical construct has been explored and discussed for a long time, and multiple instruments exist to operationalise and measure it (Wang, 2016).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Wang, 2016

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Tourism has been recognized as the number one sector in terms of online engagement.

    “It is, thus, not surprising that tourism has been recognized as the number one sector in terms of online engagement (Mack et al., 2008).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Mack et al., 2008 source

  • May combine two claims.

    Premisecitation

    The focus of previous research was on business strategy development, innovation and product development, and marketing campaigns.

    “Not surprisingly, the focus of previous research was on business strategy development, innovation and product development, and marketing campaigns (Ellion, 2007;Kuttainen et al., 2012;Pan et al., 2007).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Ellion, 2007

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Kuttainen et al., 2012

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Pan et al., 2007 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The four Vs of Big Data (volume, variety, velocity, veracity) are particularly relevant in consumer research.

    “The four Vs of Big Data, namely volume (scale), variety (different types of data), velocity (high speed, and real time), and veracity (uncertainty, and validity) are particularly relevant in consumer research (IBM, nd)”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites IBM, nd

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Travel is one of the dominant topics on social media, for example on Facebook and Twitter.

    “In addition, travel is one of the dominant topics on social media, for example on Facebook and Twitter (Neidhardt et al., 2017;Travelmail Reporter, 2013).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Neidhardt et al., 2017 source
    • cites Travelmail Reporter, 2013

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    The volume of online data relevant to the tourism context is increasing exponentially.

    “The volume of online data relevant to the tourism context is increasing exponentially.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    Methodologically, sentiment analysis represents a polarity classification problem.

    “Methodologically, sentiment analysis represents a polarity classification problem.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Interviewer bias and cultural influence in answering particular questions are known problems of survey-based approaches.

    “Interviewer bias and cultural influence in answering particular questions are other known problems of survey-based approaches (Veal, 2006).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Veal, 2006

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    Sentiment can be modelled by machines for automation, and integration across various applications.

    “Besides, sentiment can also be modelled by machines for automation, and integration across various applications (Choi et al., 2007;Rabanser & Ricci, 2005).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Choi et al., 2007 source
    • cites Rabanser & Ricci, 2005 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    TripAdvisor alone counts 350 million unique visitors per month and generates over 320 million reviews covering accommodations, restaurants, and attractions.

    “TripAdvisor alone counts 350 million unique visitors per month on their website and generates over 320 million reviews that cover accommodations, restaurants, and attractions (TripAdvisor, 2016).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites TripAdvisor, 2016

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Text summarization, and text classification along with natural language processing are earlier technologies used to facilitate information processing and data analysis.

    “Text summarization, and text classification along with natural language processing (NLP) are earlier technologies used to facilitate information processing and data analysis (Cantallops & Salvi, 2014;Ghose et al., 2012;Pan et al., 2007;Stringam & Gerdes, 2010;Xiang et al., 2015a).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Cantallops & Salvi, 2014

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Ghose et al., 2012 source
    • cites Pan et al., 2007 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Online platforms represent a two-way avenue of producing and consuming information, and 'co-creating experiences'.

    “Online platforms represent a two-way avenue of producing and consuming information, and 'co-creating experiences' (Sigala, 2016).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Sigala, 2016 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    New Internet technologies have empowered people who previously did not have a voice.

    “As a result, new Internet technologies have empowered people who previously did not have a voice (Hepburn, 2007).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Hepburn, 2007

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Despite its popularity, sentiment analysis is still in its infancy compared to earlier technologies, such as data mining and text summarization.

    “Despite its popularity, sentiment analysis is still in its infancy compared to earlier technologies, such as data mining and text summarization (Pan et al., 2007).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Pan et al., 2007 source

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Sentiment analysis can become an important tool in tourism research.

    “This review argues that sentiment analysis can become an important tool in tourism research.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Tourism is part of the 'experience economy' and those involved in the travel industry are increasingly seeking to understand the emotional and experiential elements of tourist activities.

    “Tourism is part of the 'experience economy' and those involved in the travel industry are increasingly seeking to understand the emotional and experiential elements of tourist activities (Ma, Scott, Gao & Ding, 2017).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Ma, Scott, Gao & Ding, 2017 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The most successful professional platforms in relation to travel and tourism are TripAdvisor, Expedia, VirtualTourist, and LonelyPlanet.

    “The most successful professional platforms in relation to travel and tourism are TripAdvisor, Expedia, VirtualTourist, and LonelyPlanet (Bjorkelund et al., 2012;Gretzel et al., 2007;Rabanser & Ricci, 2005).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Bjorkelund et al., 2012

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Gretzel et al., 2007

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Rabanser & Ricci, 2005 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Research on sentiment analysis goes back to the 1970s, but only recently has it received increasing attention from researchers and practitioners.

    “Whilst research on sentiment analysis goes back to the 1970, only recently it has received increasing attention from both researchers and practitioners (Brob, 2013;Pang et al., 2002).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Brob, 2013

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Pang et al., 2002 source

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    Online data related to tourism activities are generated at such velocity that they outstrip the potential of traditional (paper and pen) surveys to capture events in real-time.

    “Online data related to tourism activities are generated at such velocity that they outstrip the potential of traditional (paper and pen) surveys to capture events in real-time in order, for example, to monitor service quality and recovery.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationreasoning

    Sentiment analysis may be an indication of how data-driven research models might be of relevance to tourism research.

    “Moreover, it may be an indication of how data-driven research models might be of relevance to tourism research.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationreasoning

    Reviews may not always be subjective, therefore the binary classification needs to be extended to a ternary classification that contains a third, 'objective' category.

    “Reviews may not always be subjective, therefore, the binary classification needs to be extended to a ternary classification that contains a third, 'objective' category.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Data collection, data cleaning, mining process, and then evaluation and understanding of the results are the major steps used in most applications in relation to social media data analysis in tourism.

    “Data collection, data cleaning, mining process, and then evaluation and understanding of the results are the major steps used in most of the applications in relation to social media data analysis in tourism (Hippner & Rentzmann, 2006;Schmunk et al., 2014).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Hippner & Rentzmann, 2006 source
    • cites Schmunk et al., 2014

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Sentiment analysis, in particular in relation to customer reviews, is built on the premise that information provided through text is either subjective or objective.

    “Sentiment analysis, in particular in relation to customer reviews, is built on the premise that information provided through text (e.g., a review) is either subjective (i.e.opinionated) or objective (i.e.factual).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Feldman, 2013 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Sentiment analysis comprises a multi-step process: data retrieval, data extraction and selection, data pre-processing, feature extraction, topic detection, and data mining process.

    “Sentiment analysis comprises a multi-step process: a) data retrieval, b) data extraction and selection, c) data pre-processing, d) feature extraction, e) topic detection, and f) data mining process (e.g., Hippner & Rentzmann, 2006;Schmunk et al., 2014).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Hippner & Rentzmann, 2006 source
    • cites Schmunk et al., 2014

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Relatively less research has focused on sentence level analysis, since it is more challenging to accurately extract polarity from a small number of words compared with paragraphs and documents.

    “Relatively less research has focused on sentence level analysis, since it is more challenging to accurately extract polarity from a small number of words compared with paragraphs and documents (Brob, 2013;Choudhury, 2016;Höpken et al., 2016;Schmunk et al., 2014;Ribeiro et al., 2016).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Brob, 2013

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Choudhury, 2016
    • cites Höpken et al., 2016

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Naïve Bayes models were also adapted in an unsupervised fashion for sentiment analysis.

    “In addition, Naïve Bayes models were also adapted in an unsupervised fashion for sentiment analysis (e.g., Shimada et al., 2011).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Shimada et al., 2011 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Hu and Liu (2004) proposed a method to summarize customer reviews based on different product features.

    “Hu and Liu (2004) later proposed a method to summarize customer reviews based on different product features.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Hu and Liu (2004) source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    SANN is a combination of dictionary-and learning-based approaches that initially classifies text as either a subjective or objective review.

    “Sentiment-aware nearest neighbour model (SANN) is a combination of dictionary-and learning-based approaches that initially classifies text as either a subjective or objective review (Pappas & Popescu-Belis, 2013).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Pappas & Popescu-Belis, 2013 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    In hybrid approaches, dictionary and machine learning-based techniques can work in parallel to compute two sentiment polarities.

    “In hybrid approaches, dictionary and machine learning-based techniques can work in parallel to compute two sentiment polarities.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    Machine learning methods are further categorized into supervised and unsupervised approaches.

    “Machine learning methods are further categorized into supervised and unsupervised approaches.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Topic detection research dates back to 1998 where topic identification in the context of broadcast news was studied.

    “Topic detection research dates back to 1998 where topic identification in the context of broadcast news was studied (Allan et al., 1998).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Allan et al., 1998

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    One commonly used feature extraction technique is based on term occurrences, called term frequency (TF) or term frequency-invers document frequency (TF-IDF).

    “One of the commonly used feature extraction techniques is based on term occurrences, called term frequency (TF) or term frequency-invers document frequency (TF-IDF).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Pang et al., 2002 source
    • cites Hippner & Rentzmann, 2006 source
    • cites Menner et al., 2016 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Different methods were developed for dictionary-based approaches.

    “Different methods were developed for dictionary-based approaches (Bjorkelund et al., 2012;Bucur, 2015;Garcia et al. 2012;Hutto & Gilbert, 2014;Levallois, 2013).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Bjorkelund et al., 2012

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Bucur, 2015 source
    • cites Garcia et al. 2012

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    POS tagging is an important pre-processing task that generally forms a part of sentiment analysis by assigning each word a particular label.

    “POS tagging is an important pre-processing task that generally forms a part of sentiment analysis by assigning each word a particular label (e.g., noun, verb, and adjective).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The dictionary-based approach also includes a subcategory called semantic-based approach.

    “The dictionary-based approach also includes a subcategory called semantic-based approach (Tsytsarau & Palpanas, 2012).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Tsytsarau & Palpanas, 2012 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    SentiWordNet in itself, and in combination with a simplified Lesk Algorithm, was used in sentiment analysis.

    “SentiWordNet in itself (Bucur, 2015;Garcia et al. 2012), and in combination with a simplified Lesk Algorithm, was also used in sentiment analysis (Bjorkelund et al., 2012).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Bucur, 2015 source
    • cites Garcia et al. 2012
    • cites Bjorkelund et al., 2012

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    Each extracted review contains one or several sentences reflecting the reviewer's opinion.

    “Each extracted review contains one or several sentences reflecting the reviewer's opinion.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    Dictionary-based systems rely on the use of comprehensive sentiment lexicons and sets of fine-tuned rules.

    “Dictionary-based systems rely on the use of comprehensive sentiment lexicons and sets of fine-tuned rules.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Premisenone given

    Creating a lexicon is a one-time effort and can be used forever and often across different domains.

    “Moreover, creating a lexicon is a one-time effort and can be used forever and often across different domains.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The drawback of error promotion may be mitigated by using a backward feedback from the current to the earlier steps.

    “This drawback may be mitigated by using a backward feedback from the current to the earlier steps.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    Trends from high-intensity features can be particularly useful when visualised (e.g., on a map) to potential customers.

    “Such trends can be particularly useful when visualised (e.g., on a map) to potential customers (Bjorkelund et al., 2012).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Bjorkelund et al., 2012

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Previous research established that features with high intensity over different time periods can be useful to detect abnormal changes in hotel reviews and to analyse the reasons for these changes.

    “Previous research established that features with high intensity over different time periods can be useful to detect abnormal changes in hotel reviews, and to analyse the reasons for these changes.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Bjorkelund et al., 2012

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    The overall prediction performance of sentiment analysis methods can still be improved in both 2-class and 3-class sentiment analysis, but particularly in 3-class approaches.

    “Related to the issue of bias, the review shows that the overall prediction performance of the methods can still be improved in both 2-class and 3-class sentiment analysis, but particularly in 3-class approaches.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    Because these methods have provided reasonably good performances on Twitter data, and their lexicons also contain tourism related words and emoticons, they are applicable to the tourism domain.

    “As these methods have provided reasonably good performances on Twitter data, and their lexicons also contain tourism related words and emoticons, they are applicable to the tourism domain.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Due to the difficulty of detecting and finding implicit aspects in reviews, aspect-oriented sentiment analysis remains challenging.

    “Due to the difficulty of detecting and finding implicit aspects in reviews, aspect-oriented sentiment analysis remains challenging.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationcitation

    One reason for better positive classification might be the existence of a larger number of positive texts and lexicons' bias towards positivity, as human language is inherently biased towards positivity.

    “One reason might be the existence of a larger number of positive texts and lexicons' bias towards positivity, as human language is inherently biased towards positivity (Dodds et al., 2015).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Dodds et al., 2015 source

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    Combining sentiment scores with other data, such as information on transport, weather, the environment, special events, crises, and other destination components, may give rise to finding patterns that could not been seen and understood before.

    “Combining sentiment scores with other data, such as information on transport, weather, the environment, special events, crises, and other destination components may give rise to finding patterns that could not been seen and understood before.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    To provide a negative evaluation of the luggage delivery process in the example 'it took an hour time to receive our luggage!!!', common sense knowledge is required to interpret that an hour is not acceptable for luggage delivery.

    “To provide a negative evaluation of the luggage delivery process in the second example, common sense knowledge is required to interpret that an hour is not acceptable for luggage delivery.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    The integration of several types of data has great potential for generating future insights at scales not seen before.

    “In particular, the integration of several types of data has great potential for generating future insights at scales not seen before.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Premisenone given

    Analysing negation in reviews is semantically a complex task.

    “Moreover, analysing the negation in reviews is semantically a complex task.”
    Quote in the paperNot a substantial claim
84 of 84

  • The paper relies on this without support.

    Tourism content shared through social media has become a very influential information source that impacts tourism in many ways.

    “Tourism content shared through social media has become a very influential information source that impacts tourism in many ways.”

  • The paper relies on this without support.

    The volume of data on the Internet has reached a level that makes manual processing almost impossible, demanding new analytical approaches.

    “However, the volume of data on the Internet has reached a level that makes manual processing almost impossible, demanding new analytical approaches.”

  • The paper relies on this without support.

    In tourism, a service-based industry that relies on positive customer emotions and feedback, the concept of visitor satisfaction is of critical importance.

    “In the context of tourism, a service-based industry that relies on positive customer emotions and feedback, the concept of visitor satisfaction is of critical importance.”

  • The paper relies on this without support.

    Online social media, travel professional websites and platforms, and blogs present inexpensive means to gather rich, authentic, and unsolicited data on travellers' opinions.

    “Importantly, online social media, travel professional websites and platforms, and blogs present inexpensive means to gather rich, authentic, and unsolicited data on travellers' opinions.”

  • The paper relies on this without support.

    To analyse large data volumes more effectively, the demand for automatic multi-aspect algorithmic and machine-operated systems is increasing.

    “To analyse large data volumes more effectively, the demand for automatic multi-aspect algorithmic and machine-operated systems is increasing.”

  • The paper relies on this without support.

    Online data related to tourism activities are generated at such velocity that they outstrip the potential of traditional (paper and pen) surveys to capture events in real-time.

    “Online data related to tourism activities are generated at such velocity that they outstrip the potential of traditional (paper and pen) surveys to capture events in real-time in order, for example, to monitor service quality and recovery.”

  • The paper relies on this without support.

    Twitter's 140-character limit makes it suitable for mostly sentence-level sentiment analysis.

    “Twitter, for instance, allows tweets of up to 140 characters in length, lending itself to a mostly sentence-level sentiment analysis.”

  • The paper relies on this without support.

    Features derived based on a sentiment lexicon improve sentiment analysis results.

    “Features derived based on a sentiment lexicon improve sentiment analysis results.”

  • The paper relies on this without support.

    The use of a simple Bag of Words representation for sentiment analysis provides less favourable results compared to traditional topic classification, as semantic information needs also to be modelled by BoW.

    “It has further been shown that the use of a simple Bag of Words (BoW) representation for sentiment analysis provides less favourable results compared to traditional topic classification, as in sentiment analysis, semantic information needs also to be modelled by BoW, which is a difficult task.”

  • The paper relies on this without support.

    Considering higher order N-grams and complex linguistic features are helpful for polarity classification and can improve the results significantly.

    “Considering higher order N-grams and complex linguistic features are helpful for polarity classification and can improve the results significantly.”

  • The paper relies on this without support.

    The use of higher order N-grams and complex linguistic features is beneficial when large corpora are available for training the models.

    “However, the use of higher order N-grams and complex linguistic features is beneficial when large corpora are available for training the models.”

  • The paper relies on this without support.

    When using smaller corpora, a feature selection step is necessary to obtain satisfactory sentiment analysis results.

    “When using smaller corpora, a feature selection step is necessary to obtain satisfactory sentiment analysis results.”

  • The paper relies on this without support.

    Most travel agencies and hotel booking service providers employ scalar ratings to rate users' reviews, for example, scores between 1 and 5 stars.

    “In relation to the kind of data used for sentiment analysis in tourism, we noted that most of the travel agencies and hotel booking service providers employ scalar ratings to rate users' reviews, for example, scores between 1 and 5 stars.”

  • The paper relies on this without support.

    User-provided scores function as weakly labelled data, and can improve the classification accuracy, as well as help to verify polarity.

    “However, from an analytical perspective, the user-provided scores function as weakly labelled data, and can improve the classification accuracy, as well as help to verify polarity.”
14 of 14

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

  • Ganu review categories exceed the annotated total

    Minor

    The three reported category counts sum to 4,400, but the table lists only 3,400 annotated reviews.

    Reported 3400 · recomputed 4400

  • Kasper and Vela category counts fall short of the annotated total

    Minor

    The positive, negative, and neutral counts sum to 4,361, leaving 431 of the stated 4,792 annotated reviews unaccounted for.

    Reported 4792 · recomputed 4361

2 of 2

Cites

Flagged first, then by citations

Research works printed but not linked: 8

  1. Allan J., Carbonell J. G., Doddington G., Yamron J., Yang Y. 1998. “Topic Detection and Tracking Pilot Study Final Report.” In Proceedings of the Broadcast News Transcription and Understanding Workshop, 194-218. Burlington, MA: Morgan Kaufman.
  2. Chua A. Y. K., Banerjee S. 2013. “Reliability of Reviews on the Internet: The Case of TripAdvisor.” In Proceedings of the World Congress on Engineering and Computer Science, 453-57. New York: Springer.
  3. Claster W. B., Dinh Q. H., Cooper M. 2010. “Naïve Bayes and Unsupervised Artificial Neural Nets for Cancun Tourism Social Media Data Analysis.” In Proceedings of the Second World Congress on Nature and Biologically Inspired Computing, 158-63. New York: IEEE.
  4. Hepburn C. 2007. “Web 2.0 for the Tourism and Travel Industry: A White Paper.” ICT-Technical Reports Laboratory for Intelligent Systems in Tourism. http://hdl.handle.net/123456789/390.
  5. Hippner H., Rentzmann R. 2006. Text Mining. Ingolstadt: Springer-Verlag.
  6. Kasper W., Vela M. 2011. “Sentiment Analysis for Hotel Reviews.” In Proceedings of the Computational Linguistics-Applications Conference, 45-52. New York: IEEE.
  7. Kuttainen C. Lexhagen M. Fuchs M. Höpken W. 2012. “Social Media Monitoring and Analysis of a Swedish Tourism Destination.” In Proceedings of the 2nd Conference on Advances in Hospitality and Tourism Marketing and Management. http://miun.diva-portal.org/smash/record.jsf?pid=diva2%3A577845&dswid=-4159.
  8. Sharma D., Kulshreshtha A., Paygude P. 2015. “Tourview: Sentiment Based Analysis on Tourist Domain.” International Journal of Computer Science and Information Technologies 6 (3): 2318-20.

Cited by

Sorted by citations

Related

Authors
Alaei, A. R., Becken, S., & Stantic, B.
Affiliations
  • School of Information and Communication Technology, Griffith University, Australia
  • Griffith University (AU)
ORCID
Becken, Susanne
Venue
Journal of Travel Research
Publisher
SAGE Publications
Published
2017-12-14
Volume
58
Issue
2
Pages
175-191
ISSN
  • 0047-2875 (ISSN-L)
  • 1552-6763
DOI
10.1177/0047287517747753
OpenAlex
W2774008574
Semantic Scholar
5bc511aa30f72720260d792e57537379fb04c395
Cited by
622
References
75
Citations by year
2026: 52 · 2025: 67 · 2024: 85 · 2023: 88 · 2022: 105 · 2021: 92 · 2020: 85 · 2019: 38 · 2018: 10
FWCI
21.71
Type
journal article
Language
en
Open access
Licence
journals.sagepub.com/page/policies/text-and-data-mining-license (tdm)
Locations
  • Journal of Travel Research (publishedVersion)
  • Griffith Research Online (submittedVersion)
  • Cross Connect (Southern Cross University) (submittedVersion)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Sentiment Analysis and Opinion Mining
  • Social Sciences › Sociology and Political Science › Diverse Aspects of Tourism Research
  • Social Sciences › Sociology and Political Science › Digital Marketing and Social Media
Keywords
sentiment analysis, tourism, big data, social media, online reviews, tourist experience, computational text analysis, evaluation metrics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-11.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-11.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review