This document describes a five-filter approach for providing targeted feedback on scientific writing. The five filters are accuracy, brevity, clarity, objectivity, and formality. Examples of errors in each category are given from a corpus of graduate theses. An online error detector tool is proposed that would allow writers to input text and receive rule-based, genre-specific feedback on common errors flagged with codes. The tool is intended to help both students by providing accessible, actionable feedback and teachers by reducing time spent on repetitive error correction.

1. John Blake
Center for Language Research, University of Aizu
Responding to scientific writing
using the five-filters approach
jb11.org

2. Overview
02
1. Background: responding to scientific writing
2. Pedagogic literature: survey and analysis
3. Five filters: description
4. Practical applications:
A. No-tech: paper-based
B. Low-tech: word processor-based
C. High-tech web-based

3. 1. Background:
Teacher or student?
03
Students
Write thesis (i.e. short research article) or don`t graduate
• Reluctant to ask for help
• do not read feedback (e.g. thesis morphed)
• misinterpret feedback
• cannot understand feedback. (e.g. illegible, too pithy)
Teachers
Busy or lazy
• Time eater (one-many ratio)
• Predictable surface-level mistakes
• Expectations of “correction”
• Intended vs. perceived message
• Lack of empirical evidence

4. 1: Background: How to respond
Some of the choices
• product-focus vs. process-focus
• positive aspects vs. negative aspects Sugaring the pill (Hyland & Hyland, 2001).
• grammar vs. style - Teachers tend to focus on errors (Lee, 2008).
• all vs. some
• first instance vs. each instance
• teacher vs. peer vs. self
• text vs. audio vs. video
• code vs phrases vs. full sentences
• discontinuous vs. continuous
• stand-alone vs. consultation
• group vs. individual
• open (all students) vs. closed (only one student)
• direct vs. indirect
• automated vs. manual
• same-day vs. next-day vs. delayed
• actionable vs. not actionable
• Immediately better vs. long-term focused
04
Hyland, F. & Hyland, K. (2001).
Sugaring the pill: Praise and
criticism in written feedback.
Journal of Second Language
Writing 10, 185-212.
Lee, I. (2008). Understanding
teachers` written feedback
practices in Hong Kong
secondary classrooms. Journal of
Second Language Writing, 17 (2),
69-85.

5. 1. Background: Errors
Brown , H. (2000). Principles of Language Learning and Teaching. New Jersey: Prentice Hall.
Edge, J. (1990). Mistakes and correction. Harlow: Longman.
Orr, T., & Yamazaki, A. K. (2004). Twenty problems frequently found in English research papers
authored by Japanese researchers. In Professional Communication Conference Proceedings
International (pp. 23-35).
Selinker, L. (1972). Interlanguage. International Review of Applied Linguistics in Language
Teaching, 10(3), 209-231
Source
• Intralingual vs. interlingual errors (Selinker, 1972; Brown, 2000)
• Accidental slips, ingrained errors vs. attempts (Edge, 1990)
• Learner-induced vs. teacher-induced
Form and frequency
• Lexical, grammatical vs. discoursal
• Grammatical category (Orr & Yamazaki, 2004)
Effect
• Intrusive vs. non-intrusive errors
• Errors that lead to rejection/failure vs. other errors  My focus
05

6. 1: Background: Reasons for rejection
Bordage, 2001; McKercher et al, 2007; Pierson, 2004; Thrower,
2012 and others report the main reasons as:
• Unoriginal
• Unimportant
• Flawed (method, analysis, etc.)
• Poor language  My focus
Bordage G. (2001). Reasons reviewers reject and accept manuscripts: the strengths
and weaknesses in medical education reports. Acad Med, 76(9), 889–896
McKercher B, Law R, Weber K, Song H, Hsu C (2007). Why referees reject
manuscripts. Journal of Hospitality & Tourism Research, 31(4): 455-470
Pierson D.J. (2004). The top 10 reasons why manuscripts are not accepted for
publication. Respiratory Care, 49(10): 1246-52.
Thrower, P. (2012). Eight reasons I rejected your article. Elsevier connect.06

7. 2. Pedagogic literature
King, N. (2004). Using templates in the thematic analysis of text. In C.Cassell & G. Symon (Eds.),
Essential guide to qualitative methods in organizational research (pp. 256–270). London: Sage.
Method
• Survey of all English-language books on scientific writing provided
in national research institution for researchers
• Survey of top ten web results using keywords provided by Japanese
researcher
• Used template analysis (King, 2004) and coded advice according to
content.
• Refined code during analysis (e.g. tense  grammar  accuracy)
cf. Grounded approach – Template analysis – Content analysis
Results
• 15 English language books on scientific or research writing
• 5 catch-all groups
07

8. Type Typical problem areas
Accuracy* mistakes in facts, meaning, grammar, usage
and spelling
Brevity* too many words to say something simple
Clarity* vague or ambiguous terms
Objectivity overly subjective terms
Formality abbreviations, contractions and informal
terms
3. Five filters
* Initial coding used only three types.
08

9. Type Example errors
Accuracy The population of Japan is 12,734,100 [1]
Brevity …providing the user with various XXX and
asking him/her to...
Clarity Referring to Smith [10], Jones notes that
he…
Objectivity We are confident that XXX will become…
Formality A bunch of IT engineers collaborated and
launched…
3. Five filters: example errors
09

10. 4. Practical applications
No-tech
• Highlighter pens
Low-tech
• Categorize/classify errors
High-tech
• Online genre-specific error detector
10

11. 4A. No-tech
11

12. 4B. Low-tech
12
Highlight error, number error and provide feedback sheet

13. 4C. High-tech: Online error detector
13 Common Error Detector (Morrall, 2000-17, PolyU)

14. Proposal:
Rule-based pattern matching
1. Writer inputs text.
2. Text is searched.
3. Errors are identified.
4. Actionable feedback is displayed.
5. Writer acts on feedback.
Specific genre with high generic integrity (Bhatia, 1993)
• can target to user errors
• can rule out unsuitable phraseologies unlike MS and
Grammarly , e.g. There happened (to be a solution).
14

15. Method
Corpus collection
629 graduation theses
(2014 – 2017)
Error collection and
matching
Identify, classify and
code
Interface
Design, develop, deploy
and evaluate
15
bbetweenW+(?:w+W+){1,2}?tob/gi

16. Online writing tool: Corpus-based error detector
16

17. Accuracy errors
1. The population of Japan is 12,734,100 [1].
2. There are two types of… First,.. Second, ..Third,..
3. All students …
4. XXX will play a key factor in the near future.
5. form XX to YY
6. p < 0.5 cf. (p < 0.05) cf. (p = 0.03)
17

18. Accuracy errors
1. Factual errors related to world knowledge
2. Factual errors related to research topic
3. Overgeneralization errors
4. Overly bold claims
5. Spelling and grammar errors, esp. LaTeX users
6. Statistical errors
18

19. Brevity errors
1. The concept that was chosen as the primary
focus of this research is XXX
2. ..providing the user with various XXX and asking
him/her to XXX.
3. We analyze XXX regarding the XXX qualities, XXX
qualities and XXX qualities.
c

20. Brevity errors
1. Using multiple vague words
2. Redundant words
3. Repeated words
d

21. Clarity errors
1. XXX is something which is XXX from XXX of
somewhere of, is something which XXX
2. Referring to Smith [10], Jones notes that he…
3. XXX found two AAA and one BBB, which CCC
4. The journal plans to publish this paper were just a
rumour*
*not in corpus
21

22. Clarity errors
1. Vague expressions
2. Lexical ambiguity
3. Referential ambiguity
4. Syntactic ambiguity
5. Garden path sentences
22

23. 1. We are confident that XXX will become XXXX
2. We are pleased to announce that XXX
3. …such as services to your XXX, to your XXX, and
to XXX.
Objectivity errors
* ‘taming’ one’s subjectivity (Peshkin,1988)
Peshkin, A. (1988). In search of subjectivity. One`s own.
Educational Researcher, 17 (7), 17-21.23

24. Objectivity errors
1. Focus on people & feelings, not things & ideas
2. Emotive wording
3. Excessive personalization, e.g. use of pronouns
24

25. 1. To be more precise, act doesn’t directly cause
the effect (E).
2. This is the RQ of this paper.
3. They launched the website right after the
earthquake…
Formality errors
25

26. Formality errors
1. Contractions
2. Abbreviations
3. Slang
4. Informal terms
5. Rhetorical questions
26

27. Online writing tool: Corpus-based error detector
27
Emoticons
• Less intrusive
• Key given outside text
• Reduces repetition of text
Alphanumeric code
• Enables specific feedback
e.g. C2-1, C2-2, C2-3
• Key given outside text
• Reduces repetition

28. Conclusion
Benefit for students
• Bitesize actionable filters
• Great for multiple drafts
• Online tool: Accessible anytime anywhere
Benefit for teachers
• Reduces repetitive “correction”
• Time-saving so can focus on deeper
learning (or your own research)
28

29. http://web-ext.u-aizu.ac.jp/~jblake/ErrorDetector.html
jb11.org
jblake@u-aizu.ac.jp
Any questions, comments
or suggestions?