Understanding the Relevance Feature.

 

Relevance – What it is, How it Works, and Why It Matters

What is “Relevance” in Mary?

• Every fact (event line) that Mary extracts from your documents gets a relevance score:

  Very Low · Low · Medium · High · Very High.

• Mary checks every event against your brief: Mary gauges how strongly each event is associated with the names, dates, places, and issues you listed. The stronger the association, the higher the event climbs on the relevance scale.

How does Mary work out the relevance score?

1. Reads your Summary

   The Overview, Key Issues, People, Entities, and Additional Notes act as Mary’s “brief.”

2. Compares each fact to the brief

   If a fact shares key names, dates, medical terms, or legal concepts, its score goes up.

3. Ranks and labels

   Facts that have strong association to your Key Issues land in High or Very High; side-topics sink to Low or Very Low.

(Think of it as a smart filter that keeps your spotlight on what truly drives liability, quantum, or strategy.)

Why do we use Relevance?

• Cuts noise – Large chronologies shrink to the few facts that matter, so you read less and decide faster.
• Guides review – Junior staff can focus on High-score events first, knowing they align with the case theory.
• Speeds search & export – Filtering by score lets you build briefs, or statements in minutes.
• Increases Accuracy - Each time you tighten the Summary or adjust scores manually, Mary learns and future results get sharper.

Bottom line: A clear Summary and a good key document give Mary the context it needs to rank facts accurately—saving you time and spotlighting the evidence that wins cases.

How Mary Generates a Matter Summary

1. Mary looks for your Key Document

   • During upload, you’re invited to mark one or more files as Key Document.
   • These files act as Mary’s starting brief to generate the matter summary.
   • Scroll to the section titled, "picking a key document" to select the best key document.

2. Mary reads the Key Document front-to-back

   • It scans for the who, what, when, where, and why of the dispute.
   • Names, dates, places, injuries, contract terms—anything that frames the case—are pulled out.

3. Mary fills five Summary windows

   Summary part	What Mary writes there
   Overview	One-paragraph snapshot of the dispute.
   Key Issues	Short list of the main questions or allegations.
   People & Parties	Key individuals and their roles.
   Entities	Organisations that create important records.
   Additional Notes	Extra hints—low-relevance themes, open questions, special context.

4. Mary scores every fact against that Summary

   Every fact in the summary is used to form the criteria used to generate the relevance of each event in the chronology.
   The clearer the Summary, the sharper the relevance ratings you see in the chronology.

5. What if no Key Document is picked?

   • Mary still builds the chronology, but leaves the Summary blank.
   • Have a read of "how to generate a matter summary from scratch" for more context on how to do this.

6. You stay in control

   • Edit any Summary section to tighten wording or add context.
   • Each save triggers a quick re-scoring, so relevance stays aligned with your case theory.

A precise Key Document plus a quick polish of the Summary gives Mary everything it needs to rank the facts that win your case.

Picking a Key Document

A key document is the single best file (or small set of files) that explains the case in a nutshell. Mary builds its first Summary from whatever you choose here, so a good pick makes everything else—event tagging, relevance scores, searching—work better.

1 What makes a good key document?

What it should be	Why it helps
Central	Describes the heart of the dispute—how the injury or breach happened and who is responsible.
Clear and focused	Lays out the story without pages of side issues.
Reliable source	Written or signed by someone who knows the facts (client, insurer, doctor, court).
Well organised	Uses headings, dates, and names so Mary can pull out events and parties without guesswork.
Rich in timeline details	Gives dates and sequence of events that kick-start the chronology.

---

2 Quick checklist before you pick

1. Does it state the main allegation in plain words?
2. Does it name the key people and organisations?
3. Does it give the important dates?
4. Could a new lawyer understand the case from this document alone?
5. Is it short enough to stay on point (no thick annexures full of unrelated history)?

If you can answer “yes” to all five, you’ve found a strong key document.

---

3 Good everyday examples

Type of matter	Good key documents
Motor-vehicle injury	• Notice of Accident Claim form from the road insurer • Insurer’s liability decision letter
Workers’ compensation	• Worker’s claim form and insurer’s first response • Employer’s incident report
Medical negligence	• Formal complaint letter to hospital • Specialist report summarising the injury and alleged mistakes
General personal injury	• Client’s initial instruction letter describing the accident • Detailed case plan or resolution plan
Commercial / contract	• Statement of Claim filed in court • The signed contract or policy at the centre of the dispute
Employment	• Letter of termination setting out reasons • Application filed with the Fair Work Commission
Any matter	• Current file note, case plan, or resolution plan that sums up facts, issues, and next steps

---

4 Using more than one key document

• One document is usually plenty if it tells both the story and the damages.
• Two or three can help when no single file covers everything—for example, a claim form for facts and a case plan for strategy.
• More than three often adds noise instead of clarity, so choose sparingly.

---

5 Documents that usually don’t work well

Examples	Why they miss the mark
One blood-test result, a single X-ray report	Too narrow
Entire GP history, full payroll records, bundle of routine emails	Too broad or administrative
Blank pleading template, half-filled instruction sheet	Not enough facts

---

6 Still unsure?

Pick the most recent document that was written to brief an outsider—for example, a claim form sent to an insurer or a pleading filed in court. These documents are designed to capture the essentials and are perfect fuel for Mary.

A thoughtful choice here saves time later trimming the Summary and fixing relevance mistakes.

You can always edit the summary by clicking the edit button at the top of the platform to add or remove any details necessary. Once, you click "save changes", these will re-generate the relevance scores.

Section	Purpose	What to include	What not to include	Hot Tip
Overview	Give Mary the “elevator pitch” of the matter.	- 1–3 sentences covering who (claimant & respondent), what (alleged wrong), where/when (event date & place), and why it matters (core allegation).	- Procedural history, full chronology, lawyer reflections.	Test: can a new lawyer grasp the dispute in 15 seconds?
Key Issues	Drive relevance scoring. Mary weighs every fact against these.	- 3–6 short bullets phrased as concrete questions or assertions (e.g., “Was ACL reconstruction clinically necessary?” “Post-op infection management”). - Add qualifiers like “highly relevant” only when you mean it.	-  Generic goals (“ensure justice”), duplicate wording, low-stakes side issues.	If an event doesn’t touch a key issue, it should land in Low/Very Low.
People & Parties	Anchor names Mary should hunt for.	-  Individuals whose actions or records could change liability/quantum: surgeons, referrers, key witnesses. - Role tag: “Dr Pat Halliley – Anaesthetist”.	- Every nurse on the ward, admin staff, plaintiff’s lawyers (unless their correspondence is itself evidence).	Max ~8 names; combine roles if needed (“Radiology team at XYZ”).
Entities	Non-person organisations that show up in documents.	- Hospitals/clinics, imaging providers, employers, insurers, police station, product manufacturers. - Exact legal or trading name if possible.	- Broad descriptors (“local chemist”), places with zero document trail.	Think: places that generate records you expect to review.
Additional Notes	Add additional context to ensure accuracy of relevance, specifically low relevance tasks.	-  Known irrelevant themes (“All blood‐test results = Low relevance”).	- Substantive arguments, duplicated facts, full email chains.