Automatic identification of knowledge‐transforming content in argument essays developed from multiple sources

May combine two claims.

Eight linguistic features were important predictors of whether an evidential sentence tells or transforms source knowledge: relative position of an evidential sentence in a paragraph, absolute distance between an evidential sentence and the most recent argument, incidence of low-and high-accessibility anaphoric devices, incidence of rhetorical connectives that indicate reasoning, content-word over

“Eight linguistic features were important predictors of whether an evidential sentence tells or transforms source knowledge: relative position of an evidential sentence in a paragraph, absolute distance between an evidential sentence and the most recent argument, incidence of low-and high-accessibility anaphoric devices, incidence of rhetorical connectives that indicate reasoning, content-word over”

• AssumptionRests on an assumption: The 22 linguistic features adequately capture the cognitive processes of knowledge transforming.

May combine two claims.

For Writing mode, coders struggled to discriminate synthesis from analysis, and analysis from comprehension.

“In addition, for Writing mode, coders struggled to discriminate synthesis from analysis, and analysis from comprehension.”

May combine two claims.

Among four supervised machine learning algorithms, logistic regression had low prediction accuracy (nearly .57), while the other algorithms performed similarly (accuracy between .71 and .73).

“Except logistic regression which yielded low prediction accuracy (nearly .57)other algorithms performed similarly on the study dataset (prediction accuracy between .71 and .73).”

May combine two claims.

In the content problem space, three linguistic features are important predictors of knowledge transforming: content-word overlap between evidential sentence and source text, semantic overlap between evidential sentence and preceding argument, and semantic overlap between evidential sentence and source text.

“In the content problem space, three linguistic features are important predictors of knowledge transforming: content-word overlap between evidential sentence and source text, semantic overlap between evidential sentence and preceding argument, and semantic overlap between evidential sentence and source text.”

May combine two claims.

The probabilities for Pos_Arg_Ev < 0 could not be considered reliable because only 10 such sentences were found.

“However, as I found only 10 of those sentences in the entire dataset, the probabilities when Pos_Arg_Ev < 0 could not be considered reliable.”

May combine two claims.

Evidential sentences with evidence-argument semantic overlap (Sol_A) between 10% and 70% are moderately likely to transform source knowledge, and this likelihood drops outside that interval.

“Further, the sentences having evidence-argument semantic overlap (Sol_A) on the interval 10% -70% are moderately likely to transform source knowledge.However, this likelihood drops for sentences where Sol_A is outside this interval.”

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

Only 12 evidential sentences had negative semantic overlap with the argument.

“It should be also noted that only 12 evidential sentences had a negative semantic overlap with the argument.”

May combine two claims.

Claim(s) usually occupy the upper third of a paragraph, knowledge-transforming sentences most often come in the middle (usually within a 3-sentence window from claim), and their occurrence drops drastically in the lower third.

“From the structural point of view, claim(s) usually occupy the upper-third of a paragraph, knowledge-transforming sentences most often come in the middle (usually within a 3-sentence window from claim), and the occurrence of knowledge-transforming sentences drastically drops in the lower-third of paragraph.”

May combine two claims. The quote only partly states this claim.

Multicollinearity testing found correlations above 0.50 between several predictor pairs: Simultaneous and Reason r=0.87; Sequential and Causal_Con r=0.70; Sol_A and CWol_A r=0.55.

“The test for multicollinearity revealed the correlation of more than 0.50 between several pairs of variables:”

May combine two claims.

In the 38 essays, 479 evidential sentences were knowledge-telling, representing 67.85% of all evidential sentences (M=12.61, SD=7.68 per essay).

“479 sentences were categorized as knowledge-telling sentences (per essay: M=12.61,SD=7.68).This represents 67.85 % of all evidential sentences in 38 essays.”

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

The interactions between the content problem space and the rhetorical problem space in the knowledge-telling model are few and unproductive.

“The interactions between the content problem space and the rhetorical problem space in the knowledge-telling model are few and unproductive.”

• cites Bereiter and Scardamalia (1987)Not checked yet

  Not checked in the free run.

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

Most of the time knowledge-telling essays fail to fulfill what the writing task has demanded.

“However, even though knowledge-telling compositions manage to tell relevant knowledge within the domain outlined by keywords of the assignment, most of the time these essays fail to fulfill what the writing task has demanded, i.e., convincing readers why plastic materials are harmful.”

• cites Bereiter and Scardamalia (1987)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

The knowledge-transforming model consists of two problem spaces, rhetorical and content.

“The knowledge-transforming model consists of two problem spaces, rhetorical and content.”

• cites Bereiter and Scardamalia's (1987)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Developing knowledge-transforming skills in writing may help students improve their learning gains by actively building their knowledge, regardless of the domain.

“Accordingly, developing knowledge-transforming skills in writing may help students improve their learning gains by actively building their knowledge, regardless of the domain (Bereiter & Scardamalia, 1987).”

• cites Bereiter & Scardamalia, 1987Not checked yet

  Not checked in the free run.

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

The knowledge-telling model ignores making the connections among elements of content.

“Bereiter and Scardamalia (1987) posited that the knowledge-telling model ignores making the connections among elements of content.”

• cites Bereiter and Scardamalia (1987)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Knowledge transforming promotes learning.

“Because this process triggers reflective thinking while writing, the authors argued knowledge transforming promotes learning.”

• cites Bereiter & Scardamalia, 1987Not checked yet

  Not checked in the free run.

May combine two claims.

Post-secondary students often fail to paraphrase, interpret, and evaluate content in sources.

“Subsequent research following Bereiter and Scardamalia's (1987) research showed post-secondary students often failed to: paraphrase, interpret, and evaluate content in sources; establish novel associations across multiple sources; and integrate multiply-sourced information into coherent form (Stein, 1989;Dong, 1996;Petrić, 2007;Boscolo, Ariasi, Favero, & Ballarin, 2011;Aull, 2015).”

• cites Stein, 1989Not checked yet

  Not checked in the free run.

• cites Dong, 1996 sourceNot checked yet

  Not checked in the free run.

• cites Petrić, 2007Not checked yet

  Not checked in the free run.

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

These demands appear to prevent most postsecondary students from successfully engaging in knowledge transforming.

“These demands appear to prevent most postsecondary students from successfully engaging in knowledge transforming.”

• AssumptionRests on an assumption: Knowledge transforming is a highly demanding and challenging task where multiple cognitive and metacognitive processes interweave.

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

To meaningfully engage in knowledge-transforming processing, writers should engage constructive planning.

“To meaningfully engage in knowledge-transforming processing, writers should engage constructive planning (Flower et al., 1989).”

• cites Flower et al., 1989Not checked yet

  Not checked in the free run.

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

Stein (1990) identified four major groups of cognitive processes that occur in reading and writing from multiple sources: planning, monitoring, elaborating, and structuring.

“In their seminal work Stein (1990) identified four major groups of cognitive processes that occur in reading and writing from multiple sources: planning, monitoring, elaborating, and structuring.”

• cites Stein (1990) sourceNot checked yet

  Not checked in the free run.

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

Knowledge telling is a component of the knowledge transforming process.

“Knowledge telling is a component of the knowledge transforming process.”

• cites Bereiter and Scardamalia (1987)Not checked yet

  Not checked in the free run.

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

Tardy's study of initial master thesis and research paper drafts from two graduate students revealed challenges with structuring composition and evaluating source information to promote claims.

“The initial master thesis and research paper drafts written by two graduate students participating in Tardy's (2005) study revealed challenges writers faced with structuring composition and evaluating source information to promote their claims in the draft.”

• cites Tardy's (2005)Not checked yet

  Not checked in the free run.

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

To reach high interrater agreement among the three coders, coding involved three rounds of training together, coding independently, and calculating reliability.

“To reach high interrater agreement among the three coders, coding involved three rounds of train together → code independently → calculate reliability.”

Classified as probably not a substantial claim.

Stab and Gurevych found structural features were helpful in explaining differences between argumentative and non-argumentative units in text.

“Stab and Gurevych (2017) found structural features were helpful in explaining differences between argumentative (thesis statement, sub-claim, evidence) and non-argumentative units in the text.”

• cites Stab and Gurevych (2017)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

According to Bereiter and Scardamalia's writing model, students engaged in knowledge telling neglect cognitive and metacognitive operations that transform knowledge.

“According to Bereiter and Scardamalia's (1987) writing model, students engaged in knowledge telling neglect cognitive and metacognitive operations that transform knowledge.”

• cites Bereiter and Scardamalia's (1987)Not checked yet

  Not checked in the free run.

May combine two claims.

A writer's tendency to switch to knowledge telling under vague task requirements and avoid additional cognitive challenges is an explanation for the outcome, consistent with Weinstock's research.

“First, a writer's tendency was to switch to knowledge telling in the context of vague task requirements and thus avoid additional cognitive challenges, which resonates with findings from Weinstock's (2010) research.”

• cites Weinstock's (2010)Not checked yet

  Not checked in the free run.

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

Analyzing almost 2,000 citation instances, Petrić identified an overall tendency towards attribution.

“Analyzing almost 2,000 instances of citation use in written texts, Petrić identified overall students' tendency towards attribution, a descriptive rhetorical function that”

• cites Petrić (2007)Not checked yet

  Not checked in the free run.

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

The relative position of an evidential sentence in a paragraph identifies whether it appears toward the beginning, middle or end, allowing analysis of the relationship between position and writing mode.

“This information identifies sentences that appear toward the beginning, middle or end of a paragraph, allowing for the analysis of the relationship between a sentence's position and its writing mode.”

May combine two claims.

Students had issues elaborating source materials and structuring and revising written composition.

“Students seemed to have issues elaborating source materials, and structuring and revising written composition.”

• cites Boscolo, Arfé and Quarisa's (2007)Not checked yet

  Not checked in the free run.

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

The idea that a computer can offer scaffolding emerged from research on interactive learning environments.

“The idea that a computer can offer scaffolding emerged from research on interactive learning environments (see Reiser, 2004).”

• cites Reiser, 2004Not checked yet

  Not checked in the free run.

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

The Open Essayist analyzes essay drafts using NLP and generates feedback about rhetoric and prompts reflection.

“The Open Essayist (Whitelock, Twiner, Richardson, Field & Pulman, 2015) analyzes the essay draft using natural language processing (NLP) algorithms and generates feedback about rhetoric (e.g., most prominent words, representative sentences, essay's internal structure) and provides prompts that encourage the student's reflections on the draft.”

• cites Whitelock, Twiner, Richardson, Field & Pulman, 2015Not checked yet

  Not checked in the free run.

May combine two claims.

When the task is argumentative writing, this tendency leads to poorly articulated evidence and essays that lack persuasiveness.

“When the task is argumentative writing, this leads to poorly articulated evidence resulting in essays that lack persuasiveness.”

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

Cognitive linguistics offers means of analyzing linguistic choices writers made while constructing meaning in genre-specific texts.

“For instance, cognitive linguistics offers means of analyzing linguistic choices writers made while constructing meaning in genre-specific texts (e.g., argumentation).”

Classified as probably not a substantial claim.

In argumentative writing, an argument expressed and supported in a body paragraph may be considered as comprised of building blocks that create particular rhetorical moves.

“In argumentative writing, an argument expressed and supported in a body paragraph may be considered as comprised of these building blocks that create particular rhetorical moves (Tanko, 2017).”

• cites Tanko, 2017Not checked yet

  Not checked in the free run.

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

In Boscolo, Arfé and Quarisa's experiment, 52 undergraduate participants conceived academic writing as transforming knowledge from sources through elaboration and continuous revision rather than restating knowledge and providing personal points of view.

“The 52 undergraduate participants in Boscolo, Arfé and Quarisa's (2007) experiment conceived academic writing as a means of transforming knowledge from other sources through elaboration and continuous revision, rather than through restating knowledge and providing personal points of view.”

• cites Boscolo, Arfé and Quarisa's (2007)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

After separate coding, inter-rater agreement was calculated using the AC1 statistic.

“After separate coding, the inter-rater agreement was calculated using the AC1 statistic (Gwet, 2002).”

• cites Gwet, 2002Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Novice writers can gradually hone task-relevant cognitive processes and increase their ability to monitor and control those processes until feedback is no longer needed.

“In this way, novice writers can gradually hone task-relevant cognitive processes and increase their ability to monitor and control those processes to a point where feedback is no longer needed (Ericsson, 2006).”

• cites Ericsson, 2006Not checked yet

  Not checked in the free run.

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

Dong noted students' problems to paraphrase, interpret, synthesize, evaluate and integrate content from reading materials, and these skills present difficulties for both native and non-native English writers.

“The researcher noted students' problems to paraphrase, interpret, synthesize, evaluate and integrate content from reading materials, noting these skills present difficulties both for native and non-native English writers.”

• cites Dong (1996) sourceNot checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

The linguistic features selected for this study are proxies of cognitive processes a writer applied to relay features of source knowledge and create evidential sentences.

“The linguistic features selected for this study are proxies of cognitive processes a writer applied to relay features of source knowledge and create evidential sentences.”

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

The absolute distance feature combined with relative position may approximate a writer's engagement in text structuring and indirectly in constructive planning.

“This feature combined with relative position of evidential sentences in a paragraph may approximate writer's engagement in text structuring (e.g., arranging high-level propositions) and, indirectly, in constructive planning (e.g., organizing goals hierarchically).”

• AssumptionThe authors state this as an assumption: In a continuous sequence of claim sentences, the evidential sentence that follows the sequence is connected to the last claim in the sequence.

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

The novel approach in this study addresses both rhetorical and content aspects of writing from multiple sources.

“The novel approach in this study addresses both rhetorical and content aspects of writing from multiple sources.”

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

Petrić linked attribution to knowledge-telling and stated that attribution involves re-telling source content instead of building novel associations and elaborating sources with prior knowledge.

“Petrić linked attribution to the knowledge-telling mode of writing and pointed out that, by engaging in attribution, students are just re-telling what is stated in the sources instead of building novel associations and elaborating sources using prior knowledge.”

• cites Petrić (2007)Not checked yet

  Not checked in the free run.

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

Ineffective argument production could be further explained by a lack of metacognitive monitoring and control.

“Their production of ineffective arguments could be further explained by a lack of metacognitive monitoring (e.g., for writing task and goals) and control (e.g., editing and revising the draft).”

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

To develop cognitive and self-regulatory processes that characterize knowledge transforming, writers need effortful and scaffolded writing practice within an authentic learning environment.

“To develop cognitive and self-regulatory processes that characterize knowledge transforming, writers need to be engaged in effortful and scaffolded (i.e., externally supported) writing practice within an authentic learning environment (Proske, Narciss, & McNamara, 2012).”

• cites Proske, Narciss, & McNamara, 2012Not checked yet

  Not checked in the free run.

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

The AC1 method corrects agreement among raters for chance and is robust to skewed distribution of coding categories.

“This method corrects agreement among raters for the probability of chance agreement and is robust to skewed distribution of coding categories (Haley, Thomas, Petre, & De Roeck, 2008).”

• cites Gwet, 2002Not checked yet

  Not checked in the free run.

• cites Haley, Thomas, Petre, & De Roeck, 2008Not checked yet

  Not checked in the free run.

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

Escribo scaffolds students by providing informational hints on how to collect source information, plan, write and revise essays.

“Escribo (Proske et al., 2012) scaffolds students by providing informational hints (Roscoe, Varner, Crossley, & McNamara, 2013) on how to collect source information, plan, write and revise essays.”

• cites Proske et al., 2012Not checked yet

  Not checked in the free run.

• cites Roscoe, Varner, Crossley, & McNamara, 2013Not checked yet

  Not checked in the free run.

May combine two claims.

Even in unambiguous task contexts, such as argumentative writing, post-secondary students still tend to engage in knowledge telling and have difficulty strategically using cognitive processes embedded in knowledge transforming.

“However, as other studies reveal, even in contexts where the task requirements were unambiguous, such as an argumentative writing task investigated in this study, post-secondary students still tend to engage in knowledge telling, having hard times to strategically use cognitive processes embedded into knowledge transforming.”

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

Scholars in different fields could benefit from detailed methodology of cognitive linguistics.

“Scholars in different fields could benefit from detailed methodology of cognitive linguistics (Croft & Cruse, 2004;Dancygier, 2017).”

• cites Croft & Cruse, 2004Not checked yet

  Not checked in the free run.

• cites Dancygier, 2017Not checked yet

  Not checked in the free run.

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

Sanders and Schihperoord posited that text structure may provide insights into cognition applied in writing, arguing that appropriate analysis of written composition can reveal significant traces of the writer's cognitive representation.

“In line with this, Sanders and Schihperoord (2006) posited the text structure may provide insights into cognition applied in writing, arguing appropriate analysis of written composition can reveal”

• cites Sanders and Schihperoord (2006)Not checked yet

  Not checked in the free run.

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

Across different levels of higher education, students appear to struggle applying cognitive and self-regulatory strategies that underlie this writing model.

“The reviewed studies agree that, across different levels of higher education, students appear to struggle applying cognitive and self-regulatory strategies that underlie this writing model.”

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

The Web-Based Tutor lets students practice identifying elements of an argument using predefined formative feedback.

“In the Web-Based Tutor (Wolfe, Britt, Petrovic, Albrecht, & Kopp, 2009) students practice identifying elements of an argument (main claim, reason, counterargument, and response to the counterargument) using predefined formative feedback.”

• cites Wolfe, Britt, Petrovic, Albrecht, & Kopp, 2009Not checked yet

  Not checked in the free run.

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

To model linguistic properties of knowledge-telling and knowledge-transforming discourses, it is important to understand cognitive processes writers use when developing text.

“To model linguistic properties of knowledge-telling and knowledgetransforming discourses, it is important to understand cognitive processes writers use when developing text (Sanders & Spooren, 2007).”

• cites Sanders & Spooren, 2007Not checked yet

  Not checked in the free run.

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

Bereiter and Scardamalia emphasized the importance of the connection between linguistic research and cognitive models.

“Moreover, Bereiter and Scardamalia (1987) emphasized the importance of the connection between linguistic research and cognitive models.”

• cites Bereiter and Scardamalia (1987)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Ariel distinguished between high- and low-accessibility anaphoric devices.

“Ariel (1990) distinguished between high-and lowaccessibility anaphoric devices.”

• cites Ariel (1990)Not checked yet

  Not checked in the free run.

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

The XIP Dashboard supports post-secondary writers in searching for relevant source content, classifies sentences by rhetorical functions, and provides learning analytics to instructors.

“The XIP Dashboard (Simsek, Buckingham Shum, Sandor, De Liddo, & Ferguson, 2013) supports post-secondary writers in searching for relevant source content, classifies sentences according to their rhetorical functions (summarizing, background knowledge, contrasting ideas, novelty, significance, surprise, open question and generalization), and provides learning analytics to instructors.”

• cites Simsek, Buckingham Shum, Sandor, De Liddo, & Ferguson, 2013Not checked yet

  Not checked in the free run.

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

The aforementioned tools have demonstrated empirical benefits in supporting particular cognitive processes in learning contexts where reading and writing co-occur.

“Aiming to support particular cognitive processes in learning contexts where reading and writing co-occur, the aforementioned tools have demonstrated empirical benefits.”

Classified as probably not a substantial claim.

Classifying students' evidential sentences in terms of Bloom's taxonomy forms a basis for analytics to guide students toward producing knowledge-transforming texts with arguments strengthened by more thorough articulation of content with evidence.

“Thus, classifying students' evidential sentences in terms of Bloom's taxonomy forms a basis for analytics to guide students toward producing knowledge transforming texts with arguments strengthened by more thorough articulation of content with evidence.”

Classified as probably not a substantial claim.

Acquisition of domain knowledge might not be accomplished to its greatest extent when students limit engagement to knowledge telling.

“Moreover, the acquisition of the domain knowledge might not be accomplished to its greatest extent when students limit their engagement to knowledge telling (Bereiter & Scardamalia, 1987).”

• cites Bereiter & Scardamalia, 1987Not checked yet

  Not checked in the free run.

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

According to Bransford, Barclay and Franks, sentences represent informational blocks that build wholistic semantic descriptions at a high level.

“According to Bransford, Barclay and Franks (1972), sentences represent informational blocks that build wholistic semantic descriptions at a high level.”

• cites Bransford, Barclay and Franks (1972)Not checked yet

  Not checked in the free run.

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

Using Bloom's taxonomy to classify writers' evidential sentences could reflect underlying cognitive and metacognitive processes writers engage in.

“Using Bloom's taxonomy to classify writers' evidential sentences could reflect underlying cognitive and metacognitive processes writers engage in.”

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

Aull and Lancaster and Aull found that undergraduates lack text structuring strategies.

“Aull and Lancaster ( 2014) and Aull (2015 undergraduates lack text structuring strategies.”

• cites Aull and Lancaster ( 2014)Not checked yet

  Not checked in the free run.

• cites Aull (2015Not checked yet

  Not checked in the free run.

May combine two claims.

The essays were assessed as moderately informative and cohesive, and poorly integrated and organized.

“The essays were assessed as moderately informative and cohesive, and also poorly integrated and organized.”

• cites Boscolo, Arfé and Quarisa's (2007)Not checked yet

  Not checked in the free run.

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

The authors linked this problem to students' negligence in reflecting on the writing process.

“The authors linked this problem to students' negligence in reflecting on the writing process, explaining that although”

• cites Hyytinen, Löfström, and Lindblom-Ylänne's (2017)Not checked yet

  Not checked in the free run.

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

Petrić developed a coding scheme of nine rhetorical functions for source and citation use in master students' writing.

“Researching source and citation use in master students' writing, Petrić (2007) developed a coding scheme consisting of nine rhetorical functions: attribution, exemplification, further reference, statement of use, application, evaluation, establishing links between sources, comparison of one's own findings or interpretation with other sources, and other.”

• cites Petrić (2007)Not checked yet

  Not checked in the free run.

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

The model was selected using cross-validation on an 80% training sample and tested on a 20% hold-out sample.

“Cross-validation was performed on the training data (80% sample of dataset), while 20% was the hold-out data for testing the model performance, following suggestions by Haykin (1994).”

• cites Haykin (1994)Not checked yet

  Not checked in the free run.

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

While marking, the instructor may focus on the relationship between a claim and relevant evidence in the rhetorical space.

“While marking the argumentative essay, the instructor may focus on the relationship between a claim and relevant evidence in the rhetorical space (i.e., does the evidence support claim?).”

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

This pattern indicates a writer's tendency to generate text by referring to cues in already produced text while staying on topic defined by the argument.

“This indicates a writer's tendency to generate text by referring to cues in text already produced, while staying on topic defined by the argument.”

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

Many writers respond to task requirements by putting less cognitive effort into writing.

“This finding reflects the idea that many writers are interested in responding to task requirements by putting less cognitive effort into it.”

Classified as probably not a substantial claim.

Knowledge and comprehension categories represent less complex cognitive engagement.

“both categories representing less complex cognitive engagement (Agarwal, 2018)”

• cites Agarwal, 2018Not checked yet

  Not checked in the free run.

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

Low-accessibility anaphoric devices are noun phrases and indefinite articles; they contain much linguistic material and signal termination of the current topic and activation of another.

“Lowaccessibility anaphoric devices are noun phrases (e.g., bilingual children) and indefinite articles (a and an).Sanders and Spooren (2007) posit low-accessibility devices contain much linguistic material and, as such, signal a writer's termination of a current topic and activation of another topic.”

• cites Sanders and Spooren (2007)Not checked yet

  Not checked in the free run.

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

The result that content-word overlap is high for knowledge-transforming sentences may indicate that writers producing such sentences tended to engage in elaborating and structuring processes.

“Although this metric was expected to be high for knowledge-telling sentences, the result may indicate that writers producing knowledge-transforming sentences tended to be engaged in elaborating and structuring processes of knowledge transforming.”

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

Petrić (2007) found that students generally repeated source content by summarizing, paraphrasing, or quoting it directly.

“Petrić (2007) who reported an overall tendency of students to re-tell what is stated in sources by summarizing/paraphrasing or directly quoting it.”

• cites Petrić (2007)Not checked yet

  Not checked in the free run.

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

The study's finding that undergraduates generally convey rather than transform source knowledge aligns with Petrić (2007).

“This finding is in line with Petrić (2007)”

• cites Petrić (2007)Not checked yet

  Not checked in the free run.

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

A noun-phrase-rich knowledge-transforming sentence immediately following an argument could be a proxy for elaboration.

“For instance, a noun-phrase-rich knowledge-transforming sentence immediately following an argument could be a proxy for elaboration, one of the main cognitive processes in knowledge-transforming (Stein, 1989).”

• cites Stein, 1989Not checked yet

  Not checked in the free run.

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

Feedback on source content usage in argumentative essays has rarely been provided by instructors and has a meager effect on essay score.

“For this reason, feedback on source content usage in the argumentative essay has been rarely provided by the instructors and this characteristic of writing therefore has a meager effect on the essay score.”

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

Hyytinen, Löfström, and Lindblom-Ylänne (2017), in an experiment with roughly three times as many compositions as this study, found that most students across multiple higher-education levels struggled to select relevant source information, link main ideas, and assess alternative positions in sources.

“They demonstrated a vast majority of students from several levels of higher education faced challenges deciding on relevant source information, connecting main ideas and evaluating alternative positions presented in sources.”

• cites Hyytinen, Löfström, and Lindblom-Ylänne's (2017)Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

A scorer normally would not read sources a student cited to locate relevant information and assess transformation.

“A scorer normally would not read sources a student cited (unless suspecting plagiarism) to locate relevant pieces of information and assess that information was transformed in evidential sentences.”

May combine two claims.

These two types of linguistic markers are proxies for change in topic, which is one of the essential parameters distinguishing knowledge-telling from knowledge-transforming.

“These two types of linguistic markers are proxies for change in topic which is one of the essential parameters distinguishing knowledge-telling from knowledgetransforming.”

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

The opposite effect could be explained by a writer's disposition to take advantage of a chain of reasoning already provided in a source and translate it into a composition, avoiding further transformation.

“This could be explained by a writer's disposition to take the advantage of a chain of reasoning as already provided in a source article and just translate it into a composition, avoiding further transformation (e.g., by integrating that information with prior knowledge).”

May combine two claims.

The total essay score was selected for analysis because it and the evidence score correlate as much as reliability allows, making it a more conservative reference point.

“Since the total essay score and evidence score correlate as much as reliability allows, the total essay score was selected for analysis as it would be a more conservative reference point.”

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

The study's results resonate with the findings of Hyytinen, Löfström, and Lindblom-Ylänne's (2017) experiment.

“This resonates with results of Hyytinen, Löfström, and Lindblom-Ylänne's (2017) experiment sampling about three times more compositions than in this study.”

• cites Hyytinen, Löfström, and Lindblom-Ylänne's (2017)Not checked yet

  Not checked in the free run.

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

High-accessibility devices are pronouns and definite article; they contain less linguistic material and indicate continuation of the previous topic.

“High-accessibility devices are pronouns (personal and possessive) and definite article (the).They contain less linguistic material (Sanders & Spooren, 2007) and indicate a writer's continuation of the previous topic.”

• cites Sanders & Spooren, 2007Not checked yet

  Not checked in the free run.