5. Reading papers
efficiently is essential
Structured reading leads to faster
reading and better information retention
6. But I know how to read!
The average PhD student in science will read in
excess of 250 papers throughout the duration of
their project
Reading papers which are not exactly on topic or
from a different domain/field and can be boring!
lots of acronyms and jargon
Bulk of reading performed in first six months
7. Feeling Confused?
Familiarisation with new literature can be very daunting
It can range from frightening to completely bewildering
Lose all confidence in your abilities
Look at what all these really clever people are saying,
they are so smart it must be all correct
“I’ll just skip the paper and read the abstract”
It takes time, patience and practice to get right
8. Reasons for reading a
stack of literature
You have to become mini-experts in your field
make an original contribution
have to know who has done what, when and with whom
The recall of this information becomes ‘viva critical’
“have you read the paper of Dr. Xyz et al on your very topic??”
A lot of papers which you download will be irrelevant
its confusing as to what you ignore and what you need to keep
9. Reading high volumes
Try and read everything about your topic?
since “google scholar” et al, there is almost too much
information!
Be selective about what you read?
start with review papers, but be careful (why?)
focus on the major publications, but you could miss something....
Only read the papers your supervisor suggests?
who’s to say your supervisor is some kind of research god?
Develop a strategy or process for systematically reading papers
10. A paper != a textbook
I avoided this more by accident than by design!
I didn’t want to read all the papers all the way through as I got bored
Linear: start reading a paper at the abstract and don’t stop until you
hit the acknowledgements section
Passively reading papers in this style
Complete read of a paper would take a graduate student 3/4 hours
3 * 50 = 150 hours on reading alone
without doing anything else which is unrealistic
11. Adaptable strategies
depending on purpose
Initial familiarisation with a field
requires examination of the background sections of papers
chase down the citations and then the citations in the citation etc
quick look at the conclusions and future work
Looking for specific pieces of information
don’t print it out, this makes it very tempting to read the whole thing
make use of the fact that electronic documents are easily searchable
These are slightly more unusual cases, so what about the general purpose
paper read?
13. What this means
Never judge a paper by its title
the key to a paper is usually in the introduction not the title or abstract
last paragraph of introduction always states the contents of the paper
Leave the maths until the second read through
unless you are from maths that is
Skip next to the results section (if its an experimental paper)
what have they done and what was the outcome
Check their methods which can make more sense if you know the
outcome
14. What this means
Take a second look at the results and make assessment
do they match what is given in the introduction and methods?
in bad papers none of this stuff will match up properly
hard to see if you’re just passively reading
See what they conclude
do you agree?You don't have to, especially if there is no proof!
Use the background section to find out more information or
to clarify unfamiliar terms
15. Writing
When we read papers is it
recommended to do so in a linear
fashion?
When we write papers do we do this in
a linear fashion?
16.
17. The Blank Page
Starting a large document can be intimidating
Do not write in presentation order
Start with what you have done e.g. the
methods and results first; introduction and
abstract come last
Then you know what material you need to
support the work you are presenting
18. Plan Initial Structure
Getting the structure right is the cornerstone of
technical writing
Sketch out what you want to say
Best practice to never go beyond 3 levels of
subsections
2.3.1.4.1.5 ❌
2.3.1 ✅
19. Large Documents Are
Fractal
Large documents need strong narrative
Introductions and Conclusions for the
whole document
Within each section write and
introduction and a conclusion
Within each subsection do the same etc.
20. This creates signposting
Without narrative the document becomes very difficult
for the reader to manage
Repeating what you are doing and where in the
document aids navigation
Provides signposts so that the reader knows what they
have read and what they are going to read shortly
Remember:You are not writing crime novels, always give
your readers the full picture and avoid SUSPENSE….
21. Making Chapters Self
Contained
Good narrative and signposting can mean that any
given thesis chapter can be read independently
Minimise the amount of back and forward referencing
Skipping about while you are reading a document
disrupts the flow, and can be hard to do in digital
formats
Think about proximity and how reading can be a very
‘tactile process’
22. Getting started within a
chapter
Again, avoid the blank page if you can
Flesh out what paragraphs will go in what
section using bullet points
Should be able to see the overall
construction of a thesis at this point and
be able to read it through as a summary
23. Use Paragraphs
Appropriately!
This seems to be very simple, but is a common mistake
for MSc Students
Reader’s attention is held only for the first few sentences
of each paragraph- put the most important information at
the top of the paragraph
One paragraph, one subject
Check you have done this effectively by reading through
your chapter based on looking at the first line of each
paragraph only
24. Spot the difference
Because of the work performed by Pearson et al.
(2002) where cells were discovered to have high levels
of communicative connectivity, we can abstract a
model of cell-cell communication which adhere to
small world principles
We can abstract a model of cell-cell communication
which adheres to small world principles, based on the
findings presented in Pearson et al. (2002) where cells
were shown to have high levels of communicative
connectivity
25. Using ‘power sentences’
Transform the impact an individual sentence has by writing
it so the most important part of the sentence comes first
The reader pays more attention to the first few words of
a sentence
Helps you to focus your sentences and write more
convincingly
Often I write out a sentence as I would say it, which does
not necessarily make it either clear or concise
26. Hypotaxis vs Parataxis
Parataxis is a literary technique, in writing or
speaking, that favours short, simple
sentences, with the use of coordinating
rather than subordinating conjunctions. It
contrasts with syntaxis and hypotaxis.
[Wikipedia]
27. Parataxis
One sentence, one subject, one object
“Robots within a swarm must be able to
cooperate.Their communication devices are
tailored to their particular morphology and
functionality.Therefore communication across
different robotic platforms is difficult”
Very concise and understandable, but is very
turgid if this is written for a long time
28. Hypotaxis
Long, multiple clause, (frequently) long winded sentences),
with multiple parentheses and punctuation which are not
necessarily conducive to providing the clearest conveyance
of the true meaning of a sentence; although can be used
sparingly to break up the turgidity of constant parataxis.
If your sentence has multiple objects and topics, and is
more than two lines long, it probably needs to be
decomposed
This takes quite a lot of practice to achieve the correct
balance between parataxis and hypotaxis
29. Hypotaxis can confuse
“The decentralized decision-making mechanisms
found in the above examples, and others in the
natural world, offer an insight on how to design
distributed algorithms that solve complex
problems related to diverse fields, such as
optimization, multi-agent decision making, and
collective robotics.”
How can we re-write this to be a power
sentence with some elements of parataxis?
30. Using formal language,
fam.
The ‘tone’ of scientific papers is technical and formal
Write out contractions in full
Don’t -> Does not, Haven't -> have not
Avoid slang innit bro
Think from the perspective of your reader about the
clarity of your sentence to remove ambiguity
“Eats, shoots and leaves”
31. Common mistakes
Plurality: its it’s, data datum, sheep 🐑
Use vs utilise (did you really?)
Methodology vs method, effect vs affect
Raise vs rise, that or which, principle vs
principal
[The worst one] Could of vs Could have
32. Latin horrors
When you want to cite multiple paper
authors, for more than three authors we
use et al. Note its in italics and no stop
after the et
Confusion over e.g. and i.e.
e.g. Greensmith et al. (2008)
33. Using maths
appropriately
Introduce all notation with definitions either before or
directly after their usage
a = |b - c| * (c + d), where a is the measured response, b
is a constant, c is the contraction magnitude and d is the
contraction rate
Numbering equations is useful if you refer to them at a
later point
Simple expressions in line with the text, more complex
expressions should be centred like a table or figure
34.
35. Communicating with
figures
Select the appropriate visualisation technique
Never, never, ever, use a pie chart
Label all axes and provide legends (if line)
Assume the figure will be printed in black and
white
For each figure write an appropriate figure legend
36. Figure legends
A figure should also be a self contained object
You should not have to look at the text to understand
what is happening in a figure
Figure 2: Shows the speed-up of network traffic comparing
the standard technique with the novel peer-to-peer
implementation, based on measurements taken at 0.5Hz
on the JSim network simulator (Version 5.4)
No interpretation, but a clear description of what is in the
figure
37. The results section
All results must be described and stated in the results
section (they don't belong anywhere else)
Statistical analysis is presented in the results section
Do not repeat results, pick to either put into a graph or
into a table - no extraneous graphs
Keep methods/implementations out of this section
Tables like figures need a self contained legend and are
best referenced also like figures
38. Drawing conclusions
You should not introduce any new results in the conclusions, all
results go in the results section
Draw meaningful comparisons
“Our new technique X is better” - better than what, my cat’s dinner?
“We show that technique X is better than techniqueY.We believe
this is because…..”
Write down 1/3rd of your conclusions as “Future work”
The conclusion must come to some ‘conclusion’ and not just be a
summary of the paper.The summary is the ABSTRACT
39. Writing an abstract
This is a summary of the whole paper
The majority of people who download your
paper/thesis will only ever read this
A concise and comprehensive summary of the
work - avoiding abbreviations!
Include a summary of results and a conclusion
in the abstract
40.
41. Proof reading strategies
Factor in time for proof reading - and not just
checking for typos
Check your sentences, figures are complete, perhaps
read the first sentence of paragraphs
Check all reference information is present and
correct
I used speech to text to ‘listen’ to my theses while I
was reading them (also my Dad proofread them)
42. Further Reading and
Videos
Writing for Computer Science (Justin Zobel): I loved it so much I
bought 2 copies
https://www.amazon.co.uk/Writing-Computer-Science-Justin-Zobel-
ebook/dp/B00TEYLNBI/ref=sr_1_2?
ie=UTF8&qid=1487352524&sr=8-2&keywords=Justin+ZObel
Write now: http://www.harold.thimbleby.net/teaching/Writing.pdf
Simon Peyton Jones: https://www.youtube.com/watch?v=g3dkRsTqdDA
Elements of Eloquence: https://www.amazon.co.uk/Elements-Eloquence-
Perfect-English-Phrase-x/dp/1785781723/ref=sr_1_1?
ie=UTF8&qid=1487352598&sr=8-1&keywords=Elements+of+eloquenc
e