Summary compression works by making information smaller without changing the structure that gives the information its meaning.
That is harder than deleting words.
A source contains propositions, relationships, qualifications, hierarchy, examples, evidence and repetition. Compression decides which of those can be merged, abstracted or removed while preserving the model the receiver still needs.
Good compression reduces volume.
Bad compression reduces truth.
This article is the second pillar of How Summaries Fail.
The Core Idea: Compress Redundancy Before Meaning
The safest information to remove is usually redundancy.
Two sentences may make the same point in different words. Several examples may demonstrate one principle. Background may repeat an already-established distinction.
Compression should first remove duplicated informational function.
Only after redundancy is reduced should the writer consider removing distinct ideas.
Compression Is Not Deletion by Length
A mechanical strategy such as ‘keep one sentence from every paragraph’ preserves layout rather than meaning.
Some paragraphs contain the central claim. Others contain elaboration that can disappear.
The summary should inherit the source’s logic, not its page proportions.
Build a Source Model Before Compressing
Before writing the short version, identify the source architecture.
- Claim → reasons → evidence → limitation → implication.
- Problem → cause → response → result.
- Condition → process → outcome.
- Position A → position B → comparison.
- Sequence → turning point → consequence.
- Definition → mechanism → example → boundary.
Once the architecture is visible, compression can shorten each component without destroying the connections.
Compression Level 1: Remove Repetition
Repeated wording is the lowest-cost compression target.
If several sentences restate the same idea, combine them.
This keeps informational coverage while reducing surface length.
Compression Level 2: Group Parallel Details
A source may list several instances that share one category.
Instead of carrying each detail forward, the summary can name the category if the individual items do not change the decision.
For example, multiple examples of setup delay may compress into ‘setup and switching costs’.
The grouping label must still represent the original set accurately.
Compression Level 3: Replace Examples With the Rule
An example can often be removed once the principle it illustrates is understood.
But the rule should survive.
If the example is itself the evidence, counterexample or boundary condition, it may need to remain.
Compression Level 4: Abstract the Mechanism
Several concrete events may express one deeper mechanism.
A student opens files, finds notes, checks the assignment portal and changes subject context. A summary may compress these into ‘setup and context-switching costs’.
Abstraction is powerful because it saves words and increases transfer.
It is dangerous when the abstraction is too broad.
Compression Level 5: Collapse Detail Into a Range or Category
Exact values can sometimes become an order of magnitude, range or qualitative class.
This is appropriate only if exactness does not affect the next decision.
A report containing values of 48, 51 and 50 minutes may be summarised as ‘about fifty minutes’ for rough planning.
The same compression would be unacceptable if a threshold occurs at 50.
Preserve Logical Operators
Words that express logic can be shorter than the nouns they connect and more important.
- if
- unless
- only if
- because
- therefore
- despite
- although
- before
- after
- may
- must
- not
Removing one logical operator can reverse or strengthen the source.
Compression should protect the relationship first.
Preserve Direction
Cause has direction.
Dependency has direction.
Sequence has direction.
A summary that keeps the same concepts but reverses which one influences the other is wrong.
Use arrows, clauses or explicit connectives when direction matters.
Preserve Scope
‘Some students’, ‘most cases’, ‘one school’, ‘during the first month’ and ‘under timed conditions’ constrain the claim.
These phrases are tempting to remove because they look secondary.
They may be the exact words preventing overgeneralisation.
Preserve Contrast
Contrast is a compression problem because two ideas can be shortened separately and lose the opposition between them.
For example: ‘Practice improves fluency, but practice without feedback can automate error.’
Keeping both propositions without the but weakens the model.
Preserve Exception Structure
Rules often become dangerous when their exceptions disappear.
A compact representation can use a short condition instead of a long paragraph, but the exception must survive if it changes action.
Preserve Attribution Boundaries
A source may contain author claims, participant claims, expert quotations and opposing views.
Compression can merge them into one voice.
When attribution affects credibility or interpretation, use compact labels such as ‘the authors argue’, ‘participants reported’ or ‘critics contend’.
Use Hierarchical Compression
Long sources often compress best in stages.
- Full source.
- Section summaries.
- One integrated summary.
- Short brief.
- One-sentence orientation.
At each stage, decide what additional information can disappear.
This is more reliable than attempting a one-step jump from forty pages to two sentences.
Compression Ratios Should Change With Purpose
There is no universal ideal percentage.
A study note may retain more mechanism than an executive headline.
A search snippet may retain less nuance than a research abstract.
An examination summary may have a formal word limit that creates a different optimisation problem.
The ratio is subordinate to the information requirement.
Use Multiple Resolutions
One of the strongest systems keeps more than one summary resolution.
Headline
One sentence for orientation.
Brief
A paragraph or bullet set for the next decision.
Working summary
Enough detail for execution or study.
Source
The full material for verification and deeper reading.
This prevents one tiny summary from serving every purpose badly.
Compression and Vocabulary
Shorter wording can help.
But replacing a precise phrase with a shorter vague word can destroy meaning.
‘Independent verification’ is not always replaceable by ‘checking’.
‘Statistically associated’ is not always replaceable by ‘linked’ if the receiver may interpret linked causally.
Lexical compression should preserve the technical distinction.
Compression and Sentence Structure
Several simple source sentences can often become one complex sentence.
This saves space by sharing subjects, removing repetition and combining relationships.
The new sentence must remain readable.
Density is useful until the reader can no longer parse the structure.
Compression and Lists
Lists are efficient when items share a category.
They fail when sequence, causation or hierarchy matters.
A process written as an unordered bullet list can lose the fact that Step 2 depends on Step 1.
Choose a representation that preserves the source relation.
Compression and Tables
Tables can compress repeated dimensions across several objects.
They are especially useful for comparison.
But a table can hide narrative causality or uncertainty if cells are reduced to labels.
Use tables where shared criteria are genuinely the organising structure.
Compression and Diagrams
A diagram can replace paragraphs when relationships matter more than prose.
Dependency maps, process flows and concept maps can be excellent compressed representations.
The danger is false simplicity: arrows imply relationships and should be labelled carefully.
Compression and Numerical Data
Numbers can be compressed into central tendency, range, trend or category.
Choose the statistic that preserves the decision-relevant property.
An average can hide variation. A total can hide population size. A percentage can hide the denominator.
The shortest number is not always the best representation.
Compression and Uncertainty
Uncertainty can often be expressed compactly.
- about
- roughly
- range
- likely
- uncertain
- mixed evidence
- short term
- under tested conditions
Removing uncertainty saves very few words and can cost a great deal of accuracy.
Compression and Evidence
A full citation can remain in a reference list while the summary carries a short source cue.
The important issue is whether the reader can distinguish sourced fact from the summariser’s interpretation.
Traceability is developed fully in How Summary Traceability Works.
Compression and Student Notes
SETC’s Reading to Notes lesson uses progressive compression: paragraph → bullets → fewer bullets → sentence → label.
The useful part is the reverse test.
Can the learner reconstruct the larger model from the compressed form?
If not, compression went too far.
Compression and Mathematics
A formula sheet is extreme compression.
It is useful only when the learner already possesses the conditions, meanings and method-selection knowledge that the formula cannot carry.
A novice may need a richer representation.
An expert can operate from the compressed one.
This explains why identical notes can be sufficient for one learner and dangerous for another.
Compression and English
SETC’s Summary Writing system emphasises selection before paraphrase and then controlled compression.
A learner should combine overlapping points, remove examples and repetition where appropriate, and preserve the relationships that make the remaining ideas accurate.
Compression and Research
A research abstract is not merely a shorter paper.
It is a purpose-built representation containing a subset of the study’s structure.
Readers should still be able to identify the question, method at a useful level, central result and important limitation.
Compression and Operational Work
A shift summary or handover note should compress discussion aggressively but preserve state.
- What changed?
- What is complete?
- What is blocked?
- Who owns the next action?
- What must happen before the next state?
Operational summaries fail when they keep history but lose current state.
The Over-Compression Test
A summary is over-compressed when the receiver must guess a relationship that the source made explicit.
Common symptoms include vague nouns, missing subjects, unlabeled numbers, unexplained arrows, collapsed exceptions and one-sentence conclusions that no longer reveal their evidence boundary.
The Under-Compression Test
A summary is under-compressed when the receiver still has to perform the source-selection work themselves.
Symptoms include copied paragraphs, repeated examples, long quotations, multiple restatements of the same point and background that does not affect the purpose.
The Reconstruction Test
After compressing, set the source aside.
Explain the model from the summary.
Can you recover the central claim, structure, condition and implication?
Then compare with the source.
Anything reconstructed incorrectly identifies a compression failure.
The Reverse-Expansion Test
A useful training exercise is to expand a compressed note back into a fuller explanation.
If several incompatible explanations could fit the same note, the note may be too thin.
If the reconstruction is stable, the compression probably preserved enough structure.
The Compression Ladder
- Remove exact repetition.
- Merge equivalent statements.
- Group parallel examples.
- Replace examples with their governing rule.
- Abstract repeated mechanisms.
- Shorten language while preserving technical distinctions.
- Convert repeated comparisons into a table.
- Convert relationships into a diagram when useful.
- Remove secondary detail only after the model remains stable.
- Check qualifiers, attribution and uncertainty before finalising.
A Practical Summary-Compression Worksheet
- What can be merged without losing a distinct idea?
- Which examples are redundant?
- Which details can become a category?
- Which relationships must stay explicit?
- Which qualifiers prevent overstatement?
- Which exact numbers can become ranges or scale descriptions?
- Can a table or diagram preserve more structure in less space?
- Is the language shorter without becoming vaguer?
- Can the receiver reconstruct the intended model?
- Does the compressed version still fit the original scope?
Compression Follows Coverage
Compression should not decide what matters.
Coverage does that first.
If the wrong ideas enter the summary, elegant compression cannot repair the selection. Read How Summary Coverage Works.
Compression Must Remain Traceable
A compressed claim may be perfectly accurate today and difficult to verify tomorrow if its source path disappears.
The final pillar, How Summary Traceability Works, keeps compressed information connected to its evidence.
Across the eduKate Ecosystem
eduKateSG owns the general compression mechanism. SETC Summary Writing applies controlled reduction to English, while Reading to Notes trains several compression resolutions. Bukit Timah Tutor protects mathematical conditions and verification when symbolic representations become compact. eduKate Sengkang adds the reconstruction test: the learner must be able to recover and use the knowledge independently.
Sources and Further Reading
Purdue OWL — Quoting, Paraphrasing and Summarizing
Continue the Series
How Summaries Fail | Why Compression, Omission and Distortion Break Meaning
How Summary Coverage Works | Keep Decision-Critical Information
How Summary Traceability Works | Keep Claims Connected to Sources
