Stance Analysis

Background Osteoarthritis (OA) is the most commonly diagnosed joint disease in veterinary medicine, with at least 80% of the cases of lameness and joint diseases in companion animals being classified as OA. Sporting and working animals are more predisposed to develop OA since they are exposed to chronic fatigue injuries, leading to bone and muscular tissue damage and failure, resulting in clinical signs. To characterize the clinical signs and diagnostic findings of Police working dogs presenting with bilateral hip OA at the time of diagnosis. Fifty animals were evaluated with a bodyweight ≥ 15 kg, be older than two years, and without any medication or nutritional supplements for ≥ 6 weeks. Results Weight distribution, joint range of motion at flexion and extension, thigh girth, digital thermography, and radiographic signs were collected. Data from different Clinical Metrology Instruments (CMI) were collected: Canine Brief Pain Inventory, Liverpool Osteoarthritis in Dogs, Canine Orth...

Figure 1: Visualization of sentiment and stance data series discovered in social media data in our VA system StanceVis Prime: (a) a list of tracked targets of interest, corresponding data domains, and detected subjectivity categories; (b) a stacked graph representing the processed document counts and cues about detected subjectivity levels; (c) a slider controlling the threshold for displaying subjectivity cues in (b); (d) a range slider providing an overview about the complete data set; (e) bar charts representing sentiment and stance data series; (f) a document view panel; and (g) a tooltip with detailed sentiment and stance classification results for a specific document.

The automatic detection and classification of stance (e.g., certainty or agreement) in text data using natural language processing and machine-learning methods creates an opportunity to gain insight into the speakers' attitudes toward their own and other people's utterances. However, identifying stance in text presents many challenges related to training data collection and classifier training. To facilitate the entire process of training a stance classifier, we propose a visual analytics approach, called ALVA, for text data annotation and visualization. ALVA's interplay with the stance classifier follows an active learning strategy to select suitable candidate utterances for manual annotaion. Our approach supports annotation process management and provides the annotators with a clean user interface for labeling utterances with multiple stance categories. ALVA also contains a visualization method to help analysts of the annotation and training process gain a better understanding of the categories used by the annotators. The visualization uses a novel visual representation, called CatCombos, which groups individual annotation items by the combination of stance categories. Additionally, our system makes a visualization of a vector space model available that is itself based on utterances. ALVA is already being used by our domain experts in linguistics and computational linguistics to improve the understanding of stance phenomena and to build a stance classifier for applications such as social media monitoring.

Background Osteoarthritis (OA) is the most commonly diagnosed joint disease in veterinary medicine, with at least 80% of the cases of lameness and joint diseases in companion animals being classified as OA. Sporting and working animals are more predisposed to develop OA since they are exposed to chronic fatigue injuries, leading to bone and muscular tissue damage and failure, resulting in clinical signs. To characterize the clinical signs and diagnostic findings of Police working dogs presenting with bilateral hip OA at the time of diagnosis. Fifty animals were evaluated with a bodyweight ≥ 15 kg, be older than two years, and without any medication or nutritional supplements for ≥ 6 weeks. Results Weight distribution, joint range of motion at flexion and extension, thigh girth, digital thermography, and radiographic signs were collected. Data from different Clinical Metrology Instruments (CMI) were collected: Canine Brief Pain Inventory, Liverpool Osteoarthritis in Dogs, Canine Orth...

Purpose Extracting information from unstructured data becomes a challenging task for computational linguistics. Public figure’s statement attributed by journalists in a story is one type of information that can be processed into structured data. Therefore, having the knowledge base about this data will be very beneficial for further use, such as for opinion mining, claim detection and fact-checking. This study aims to understand statement extraction tasks and the models that have already been applied to formulate a framework for further study. Design/methodology/approach This paper presents a literature review from selected previous research that specifically addresses the topics of quotation extraction and quotation attribution. Research works that discuss corpus development related to quotation extraction and quotation attribution are also considered. The findings of the review will be used as a basis for proposing a framework to direct further research. Findings There are three f...

The objective of this paper is to understand how these companies are being mapped and analyzed in relation to sales, with the support of Digital Analytics tools. It was selected the top 10 automotive companies present in Brazil with the highest sales and data were collected on their pages through Digital Analytics tools during two months. It was chosen the multivariate technique of multiple linear regressions by analyzing the relation between the independent variables (collected attributes) with the dependent variable (sales). As a result, it was found that some tools have a better set of parameters that explains the sales of automakers. From the seven Digital Analytics software's observed, six-showed significance in explanatory power. This research was purely quantitative, in which the independent variables that were significant for this study could be grouped into: "Social Media" and "Not Social Media", overcoming a greater concentration of variables "...

The objective of this paper is to understand how these companies are being mapped and analyzed in relation to sales, with the support of Digital Analytics tools. It was selected the top 10 automotive companies present in Brazil with the highest sales and data were collected on their pages through Digital Analytics tools during two months. It was chosen the multivariate technique of multiple linear regressions by analyzing the relation between the independent variables (collected attributes) with the dependent variable (sales). As a result, it was found that some tools have a better set of parameters that explains the sales of automakers. From the seven Digital Analytics software's observed, six-showed significance in explanatory power. This research was purely quantitative, in which the independent variables that were significant for this study could be grouped into: "Social Media" and "Not Social Media", overcoming a greater concentration of variables "...

Language modeling and its applications are influenced by data quality and the richness of the features that are able to extract, especially for low-resource languages such as Arabic. This paper presents the first empirical effort in annotating attribution in Modern Standard Arabic (MSA). We based on the Arabic Treebank (ATB) which is a corpus that constitutes the gold standard annotation for morphological, syntactic, and discourse features as in the LADTB. The identification of attributed arguments to a source, or so called author identification, has been applied successfully in diverse systems such as information retrieval, and opinion/sentiment mining for several languages. Although lexical and syntactic features play a significant role in these applications, there is a huge demand for Arabic attribution resources in long and short text. We, therefore, insight of prior efforts for other languages, annotate several features for Arabic author identification systems: the basic elemen...

Oral Proficiency Interviews (OPIs) are widely used to measure speaking ability in a second or foreign language. The Michigan English Language Assessment Battery (MELAB) Speaking Test is an OPI used for academic and professional purposes around the world. However, little research on this or other OPIs has quantitatively compared test takers' speech with the target domains of the test. Such a comparison could be used as evidence for the validity argument (Kane, 2013) for the MELAB. In this study we use corpus-based register analysis and Multi-Dimensional (MD) analysis, investigating a large number of linguistic features to determine the extent to which the language of the MELAB is similar to conversational, academic, and professional spoken discourse, specifically nurse-patient interactions, since many of the test takers are preparing for nursing licensure. The results show that the MELAB has similarities with conversation in its use of stance, and is closely aligned with academic registers and nurse-patient interactions in the use of language for informational exchange, which provides support for the validity argument of the MELAB. However, the use of narrative features and discussion of future possibilities and suggestions are important aspects of both conversation and academic and professional registers but may be harder to evaluate through the MELAB and other similar OPIs.

The British Academic Written English (BAWE) corpus contains about 6.5 million words of proficient university student writing, categorised in terms of ‘genre families’ and distributed fairly equally across levels of study and disciplinary groupings. The corpus has been examined from various perspectives, including multidimensional analysis (MDA) (Nesi & Gardner 2012) using Biber’s 1988 dimensions. This analysis revealed clear distinctions between texts produced by first, second and final year undergraduates and Masters students, and between texts belonging to different disciplinary groupings. In terms of stance, epistemic modality was the most discriminating feature, being considered a tool for persuasion on Dimension 4. Published academic prose from the Lancaster-Oslo-Bergen (LOB) Corpus had a neutral score on Dimension 4 in Biber’s own 1988 study, but scores for BAWE assignments were negative on Dimension 4, and became increasingly so across levels of study (Table 1). Scores for Ar...

Text visualization techniques often make use of automatic text classification methods. One of such methods is stance analysis, which is concerned with detecting various aspects of the writer's attitude towards utterances expressed in the text. Existing text visualization approaches for stance classification results are usually adapted to textual data consisting of individual utterances or short messages, and they are often designed for social media or debate monitoring tasks. In this paper, we propose a visualization approach called DoSVis (Document Stance Visualization) that focuses instead on individual text documents of a larger length. DoSVis provides an overview of multiple stance categories detected by our classifier at the utterance level as well as a detailed text view annotated with classification results, thus supporting both distant and close reading tasks. We describe our approach by discussing several application scenarios involving business reports and works of literature. 1 Advances in the description and explanation of Stance in discourse using Visual and Computational Text Analytics (http://cs.lnu.se/stavicta/).

Online social media are a perfect text source for stance analysis. Stance in human communication is concerned with speaker attitudes, beliefs, feelings and opinions. Expressions of stance are associated with the speakers' view of what they are talking about and what is up for discussion and negotiation in the intersubjective exchange. Taking stance is thus crucial for the social construction of meaning. Increased knowledge of stance can be useful for many application fields such as business intelligence, security analytics, or social media monitoring. In order to process large amounts of text data for stance analyses, linguists need interactive tools to explore the textual sources as well as the processed data based on computational linguistics techniques. Both original texts and derived data are important for refining the analyses iteratively. In this work, we present a visual analytics tool for online social media text data that can be used to open up the investigation of stan...

The connections between sociolinguistic conceptions of stance and the notion of affect are explored. After an explication of sociolinguistic theory, a view of stance based on three dimensions of evaluation, investment, and alignment is articulated. It is argued that low investment stances are connected to masculinity, and through that to a wider affect of 'masculine ease.' ARTICLE HISTORY

Pakistani English as a non-native variety exhibits variation at different levels of language. Early quantitative studies on Pakistani English have compared individual linguistic features of Pakistani English with their counterparts in British English and claimed about the distinctive identity of Pakistani English as an indigenous variety. Pakistani English need to be compared at the level of register to further highlight its unique features and strengthen its distinct identity. Based on a special purpose corpus, the present research paper endeavors to investigate linguistic variation across disciplines in Pakistani academic writing as a register. Disciplinary variation is explored along with five new textual dimensions identified and labeled through the technique of Multidimensional analysis (Azher & Mehmood, 2016). The ANOVA results reveal that statistically significant differences are found among disciplines on all the new dimensions of Pakistani Academic Writing. The findings und...

On the topic of journalistic integrity, the current state of accurate, impartial news reporting has garnered much debate in context to the 2016 US Presidential Election. In pursuit of computational evaluation of news text, the statements (attributions) ascribed by media outlets to sources provide a common category of evidence on which to operate. In this paper, we develop an approach to compare partisan traits of news text attributions and apply it to characterize differences in statements ascribed to candidate, Hilary Clinton, and incumbent President, Donald Trump. In doing so, we present a model trained on over 600 in-house annotated attributions to identify each candidate with accuracy > 88%. Finally, we discuss insights from its performance for future research.

תקציר בעברית: במאמר זה מוצגת גישה חדשה לניכוי הפחת השונה באופן מהותי מהגישה הרווחת ‏במשטרי מיסוי ההכנסה השונים. ניכוי הפחת הקלסי הוא ניכוי רב־שלבי שנועד להחזיר ‏את עלות הנכס‎ ‎‏(‏‎(Recovery Cost Capital‏ לאורך תקופת הנכס. הגישה המוצעת גורסת כי ‏יש להמיר את זרם ניכויי הפחת הרב־שנתי בניכוי פחת חד־שנתי על בסיס הערך הנוכחי. ‏ערכו הכלכלי של הניכוי המוצע זהה לערכו הכלכלי של ניכוי הפחת הרב־שנתי. חד ‏שנתיותו של הניכוי המוצע אין בה, כשלעצמה, כדי להופכו להוצאת מס. סיווגו של ניכוי ‏הפחת החד־שנתי על בסיס הערך הנוכחי, כהוצאת מס או כחלק ממבנה המס הנורמטיבי, ‏הולך אחר סיווגו של ניכוי הפחת הרב־שנתי שהוון לניכוי הפחת החד־שנתי.‏ לגישה המוצעת יתרונות רבים, הן במשטר מיסוי נומינלי והן במשטר מיסוי מותאם ‏לאינפלציה, ובמרכזם: חסכון בעלויות הציות והאכיפה של ניכוי הפחת בהיותו חד־שנתי ‏ולא רב־שנתי; לאחר ניכוי הפחת המוצע עלות הנכס בספרי הנישום היא אפס‎;‎‏ אין כל צורך ‏בחישוב רווח ההון משום שכל התמורה היא רווח הון‎;‎‏ הניכוי המוצע אף כולל באופן ‏אינהרנטי מערכת התאמה לאינפלציה בשני מובנים: ראשית ניתן לוותר על המ...

This article represents a further exploration of the viability of the model of Linear Unit Discourse Analysis (LUDA) proposed by Smart (2016) through the analysis of the spoken discourse of candidates in the peer group interaction component of the Hong Kong Diploma of Secondary Education (HKDSE) speaking public examination. LUDA is a linear description of analysis following the tradition of John Sinclair and his associates (e.g., Brazil, 1995; Sinclair, 1992, 1993/2004; Sinclair & Mauranen, 2006), which purports to describe the linguistic phenomena that signal the links between the sequence of incoming language as the listener experiences it. In this article, the HKDSE candidates' discourse is compared with the discourse of a university seminar discussion from the MICASE corpus using LUDA. It is found that HKDSE candidates' discourse tends to display little in terms of organizational-interactional linear units prevalent in the MICASE data. Neither does the HKDSE candidates&#...

Follow this and additional works at: http://repository.jmls.edu/facpubs Part of the Communication Technology and New Media Commons, Comparative and Foreign Law Commons, Computer Law Commons, European Law Commons, Internet Law Commons, Marketing Law Commons, Privacy Law Commons, Science and Technology Law Commons, Science and Technology Policy Commons, Science and Technology Studies Commons, Social Media Commons, and the Technology and Innovation Commons

This article offers a linguistic analysis of interpersonal stancetaking in four argumentative term papers written in an upper-level undergraduate course in economics. Two of the papers were written by English L2 writers who experienced particular difficulty with the assignment and two by English L1 writers who received the highest grades among the forty students in the course. My analysis is guided by the question of how recurring patterns of interpersonal meanings operate to construct an argumentative stance that indexes, or not, the specific goals and expectations for writing in the course. Considered alongside interviews with the professor and the graduate student instructor (GSI) who graded and commented on the papers, my analysis draws on the Engagement framework from systemic functional linguistics (Martin & White, 2005), which has proven useful in recent years for understanding the ways both professional and student writers (Chang & Schleppegrell, 2011; Derewianka, 2009; Wu, ...

On the topic of journalistic integrity, the current state of accurate, impartial news reporting has garnered much debate in context to the 2016 US Presidential Election. In pursuit of computational evaluation of news text, the statements (attributions) ascribed by media outlets to sources provide a common category of evidence on which to operate. In this paper, we develop an approach to compare partisan traits of news text attributions and apply it to characterize differences in statements ascribed to candidate, Hilary Clinton, and incumbent President, Donald Trump. In doing so, we present a model trained on over 600 in-house annotated attributions to identify each candidate with accuracy > 88%. Finally, we discuss insights from its performance for future research.

I. INTRODUCTION Like Microsoft in the 90s, Google has been at the center of much high-profile Intellectual Property-related litigation (such as the Google Books Settlement and AdWords litigation) and technology-based privacy concerns (such as de-anonymizing Google search, Google Streetview, and the controversy over Google Buzz). However, the next wave of concern regarding Google involves web analytics. Web analytics is the measurement, collection, analysis, and reporting of Internet data for the purposes of understanding and optimizing web usage.' The concerns of web analytics use touches on issues of online user privacy, government use of personal information, and information on website user activity. The profession of web analytics has formally existed since the early 90s and Google Analytics has been available since 2005. While Google Analytics is not the sole web analytics product on the market, it is widely used by corporate, non-profit, and government organizations. The product has been reported to have a 59% market share among web analytics vendors in a 2008 study. 2 Web analytics technology has also recently become the focus of government review in both the U.S. and the E.U. In May 2009, the Center for Democracy & Technology (CDT) and the Electronic Frontier Foundation (EFF) released a joint paper, Open Recommendations for the Use of Web Measurement Tools on Federal Government Web Sites. 3 In the United States, as part of a larger commitment to a consistent technology policy, in July 2009, the Office of Management and Budget asked for comments regarding web-tracking technologies, such as cookies. 4 In June 2010, in response to the comments, the Office of Management and Budget released two highly influential documents relating to the collection of personally identifiable information through web tracking technologies on

1. The frequency of 'END UP' In a preliminary exploration of the most frequent phrasal verbs (PVs) used in the MICASE corpus of spoken academic English, we ended up with the following data: Although a more detailed analysis on the PV status of these verbs is needed, the reader is most likely struck by the presence of 'end up' and 'figure out' in the top five. This kibbitzer therefore is an attempt to explore and, if possible, explain the roles of the former phrasal verb in this academic register. We next compare 'end up' in MICASE with other corpora, producing the results in Table 2. As can be seen, 'end up' in MICASE is more than twice as frequent as in any of the other corpora. The table also shows (as ELFA is also a spoken academic corpus of non-native speakers of English) that 'end up' is very much associated with the spoken mode.

Stance voluntarism highlights the role of the will in epistemic agency, claiming that agents can control the epistemic stances they assume in forming beliefs. It claims that radical belief changes are not compelled by the evidence; they are rationally permitted choices about which epistemic stances to adopt. However, terms like "will", "choice", and "stance" play a crucial role while being left as vague notions. This paper investigates what kind of control rational agents can have over epistemic stances. I argue that whether epistemic stances are voluntary depends on what kind of stance is being assessed. Sometimes epistemic stances are taken to be evaluative attitudes about how to produce knowledge. This kind of stance is not directly controllable, since it is essentially connected to beliefs, and believing is not voluntary. But sometimes epistemic stances are taken to be styles of reasoning and modes of engagement, expressing ways of approaching the world in order to produce knowledge, which can be voluntary. Overall, this supports a formulation of stance voluntarism as a dual-systems theory of epistemic agency, where epistemic rationality is compounded by a dynamic interplay between involuntary processes of belief formation and voluntary processes of cognitive guidance.

The main aim of the present study is to build knowledge and understanding regarding the linguistic use of hedging by Malaysian speakers of English in academic spoken discourse and, more specifically, doctorate students presenting their thesis in the candidature defense. It looks not only to describe the specific nature of spoken academic hedging in Malaysia but to make inroads into identifying cultural factors that shape the way this hedging takes place. To realize these aims, a corpus-based investigative case study that used both quantitative and qualitative methods was used. Transcriptions of candidature defenses and semi-structured interviews regarding the nature of doctoral candidature defense were used to gather the data. The study found that national culture and the dynamic power distance in the educational culture surrounding the defense had an effect on the interpersonal hedging used by Malaysian speakers and the gap in hedging use between panel members and candidates. Disci...

The main aim of the present study is to build knowledge and understanding regarding the linguistic use of hedging by Malaysian speakers of English in academic spoken discourse and, more specifically, doctorate students presenting their thesis in the candidature defense. It looks not only to describe the specific nature of spoken academic hedging in Malaysia but to make inroads into identifying cultural factors that shape the way this hedging takes place. To realize these aims, a corpus-based investigative case study that used both quantitative and qualitative methods was used. Transcriptions of candidature defenses and semi-structured interviews regarding the nature of doctoral candidature defense were used to gather the data. The study found that national culture and the dynamic power distance in the educational culture surrounding the defense had an effect on the interpersonal hedging used by Malaysian speakers and the gap in hedging use between panel members and candidates. Disci...

This paper investigates the phenomenon of involvement in American tv political interviews. The dataset used to carry out the study consists of a corpus of interviews with American leading politicians aired on the most popular US Sunday morning talk shows within a time span of nine years, from 2010 to 2019. Comparisons are made with another synchronic corpus of American tv interviews coming from the entertaining talk show Charlie Rose featuring personalities who discuss different issues such as technology, science, the film industry, etc. Corpus linguistics methods are exploited to describe verbal traits associated with involvement in American political interviews, as compared to entertaining-oriented talk show interviews. Therefore, this study ascertains whether, and in which ways, the usage of linguistic involvement characterizes the genre of political interviews.

On the topic of journalistic integrity, the current state of accurate, impartial news reporting has garnered much debate in context to the 2016 US Presidential Election. In pursuit of computational evaluation of news text, the statements (attributions) ascribed by media outlets to sources provide a common category of evidence on which to operate. In this paper, we develop an approach to compare partisan traits of news text attributions and apply it to characterize differences in statements ascribed to candidate, Hilary Clinton, and incumbent President, Donald Trump. In doing so, we present a model trained on over 600 in-house annotated attributions to identify each candidate with accuracy > 88%. Finally, we discuss insights from its performance for future research.

The connections between sociolinguistic conceptions of stance and the notion of affect are explored. After an explication of sociolinguistic theory, a view of stance based on three dimensions of evaluation, investment, and alignment is articulated. It is argued that low investment stances are connected to masculinity, and through that to a wider affect of 'masculine ease.'

On-line communication has exploded over the past decade. From bulletin boards, to email to blogs, the advantages are many and using this type of instantaneous communication that is available to everyone and at any time is only limited to what you wish to say. Stance studies in language investigate how interlocutors position themselves with their utterances in a dialogic interaction. This synchronic study employs a construction of stance and stancetaking based on DuBois's evaluation -position -alignment framework (2007). More specifically this study addresses how the emoticon signifiers, termed emotionizers, :) and :( are used as positive and negative signified emotions in online posts, and how they are employed as emotional intensifiers and/or downtoners, of stance in lieu of words.

On-line communication has exploded over the past decade. From bulletin boards, to email to blogs, the advantages are many and using this type of instantaneous communication that is available to everyone and at any time is only limited to what you wish to say. Stance studies in language investigate how interlocutors position themselves with their utterances in a dialogic interaction. This synchronic study employs a construction of stance and stancetaking based on DuBois's evaluation -position -alignment framework (2007). More specifically this study addresses how the emoticon signifiers, termed emotionizers, :) and :( are used as positive and negative signified emotions in online posts, and how they are employed as emotional intensifiers and/or downtoners, of stance in lieu of words.

The MICASE Handbook complements the Michigan Corpus of Academic Spoken English (MICASE), which contains transcripts of 152 speech events recorded at the University of Michigan between 1997 and 2001. MICASE represents speech in a variety of academic settings, such as lectures, presentations, study groups and meetings, across 47 departments in four broad academic divisions. It is primarily intended as a resource for linguists and language teachers, whilst also offering social historians ''an oral and written glimpse of a place and time in history''. The corpus is available for purchase and can also be accessed via a free online interface (http://quod.lib. umich.edu/m/micase).

In English for Academic Purposes contexts, spoken language has had rather less attention than aspects such as academic writing, vocabulary, listening or grammar. This comment may be a speculation rather than a verified fact, but it is one that can confidently be made if we go by a survey of the contents of this journal. The reasons for this are varied e the place of spoken language in academic assessment might be one justification e but it is likely that the slippery nature of speech and the inherent difficulties of lack of ready access to examples of recorded oral text, ethical considerations, and problems of authenticity provide other reasons why researchers may shy away from spoken language research. Fortunately this is changing, due in part to the emergence of relevant corpora, for example, and the Michigan Corpus of Academic Spoken English (MICASE) (2007). Readers of this collection of 12 articles, edited by Rebecca Hughes, might come away stimulated to engage in studies of aspects of spoken language, and more convinced of the place of spoken language in the core curriculum of TESOL and applied linguistics. The book is relevant to teachers, testers and researchers as it presents a collection of recent articles which together constitute a survey of the 'state of play' in the field.

Research on the linguistic characteristics of university classroom discourse highlights the salience, in this register, of non-informational and subjective aspects of discourse. This dimension of classroom discourse however has not been studied systematically. Taking a corpus-based approach, this study investigates the non-informational dimension of classroom discourse, focusing on the marking of emotions, affect, and the speaker’s ‘involvement’ (Chafe 1982) in the talk of university professors, in a large corpus of American university classroom discourse. It tracks the use of involvement markers across class sessions representing three situational factors that define the university setting: academic discipline, level of instruction, and class size. Surprisingly, these factors have relatively little influence on the involvement in language, suggesting that involvement is pervasive in American classroom talk. However, involvement tends to be more common in small courses in the Humanities and Social Sciences, at the upper-division and graduate level – courses which generally favor student participation, hence interactivity in discourse. Analyses of the relationship between involvement and interactivity reveal that while interactivity tends to predict involvement, involvement is not limited to interactive discourse.

This paper aims to identify textual characteristics and differences between English textbooks in Japan and other major Asian countries (China, Korea, and Taiwan). To this end, the author first reviewed the related literature and pointed out that the English textbook research so far has dealt with a very limited scope in terms of the coverage of English textbooks as a whole. Very little has been done on the analysis of English textbooks from genre/text type perspectives. In order to fill the gap, the present study employed Biber’s multidimensional (MD) analysis in the comparison of English textbooks in Japan, China, Korea, and Taiwan. Three dimensions showing different textual characteristics were extracted and functional interpretations were given. The four countries were then compared along these dimensions. The experiment shows that there are previously unrevealed abstract differences among English textbooks in different countries.