How to Resolve Translation Memory Conflicts

Conflicting translation memory entries—identical source segments with multiple target translations—are one of the most common and frustrating problems in TM management. When a translator requests a match, the TM presents multiple options with no clear guidance on which is correct. The result is inconsistent output, wasted decision time, and declining confidence in the TM itself.

This guide walks through a systematic approach to identifying, resolving, and preventing TM conflicts. If the source and target are the same and the TM stored the pair twice, that is a write-policy problem, not this workflow: how to prevent duplicate segments. If the same concept was approved as different wording across jobs but the full source segments are not identical, that is terminology inconsistency across projects, not a TM-conflict filter.

What Causes TM Conflicts?

Conflicts arise when the same source text gets translated differently across projects or by different translators, and both translations are stored in the TM. Common causes include:

  • Multiple translators with different preferences: Translator A uses formal register; Translator B prefers conversational. Both translations get confirmed and stored.
  • Evolving terminology: Your organization adopts a new preferred term, but the old term remains in the TM alongside the new one.
  • Context-specific translations: The same source phrase appears in a technical manual and a marketing brochure, requiring different translations. Without context-aware matching, both get stored as alternatives.
  • Role-based duplicates: If your TM stores role metadata (translator, reviewer 1, reviewer 2), confirming the same segment in different roles can create duplicate entries even with identical source and target.
  • Improper TM configuration: Importing TMs without deduplication, or enabling "allow multiple translations" without a governance process, amplifies conflicts.

Over time, these conflicts accumulate. A TM that starts with 100% unique entries can degrade to 20-30% conflicting entries within a few years of active use.

Step 1: Identify Conflicting Entries

Most TM tools provide filters or queries to detect duplicates and conflicts. The exact interface varies by tool, but the logic is consistent:

In memoQ

Use the Filter for duplicates function (available in the TM editor):

  1. Open your TM in the memoQ TM editor
  2. Select Operations → Filter for duplicates
  3. Choose matching criteria:
    • Sources equal: Finds all entries with identical source text (different targets)
    • Sources equal, roles different: Finds duplicates created by different workflow roles
    • Sources and targets both equal: Pure duplicates (same source and target, useful for metadata cleanup)
  4. memoQ groups duplicate entries and lets you designate one as the "master" for each group

In Trados Studio

Use the Translation Memory Viewer:

  1. Open your TM in the TM Viewer
  2. Navigate to Edit → Filter
  3. Apply a custom filter for "Source Text" with grouping by source segment
  4. Review groups with multiple entries; sort by target text to compare translations side-by-side

Prioritize High-Impact Conflicts

Don't try to resolve every conflict immediately. Focus first on:

  • High-frequency segments: Conflicts in segments reused across many projects have the greatest impact on consistency
  • Key terminology: Product names, technical terms, and domain-specific phrases where inconsistency is most visible
  • Recent projects: Conflicts from active projects are more urgent than legacy content

Step 2: Analyze Each Conflict

For each conflicting group, compare the translations side-by-side. Ask:

Accuracy and Meaning

  • Does each translation convey the source meaning correctly?
  • Are there subtle semantic differences (e.g., formal vs. informal register, active vs. passive voice)?
  • Does one translation omit nuance or add unintended connotation?

Terminology Consistency

  • Which translation aligns with your current terminology database and style guide?
  • If neither matches current standards, both may need to be replaced

Context and Usage

  • Review the original context (if available in TM metadata). Was the translation created for a specific domain, audience, or content type?
  • Check which projects or clients used each variant. If one variant is tied to a legacy client or discontinued product line, it may be safe to archive

Recency and Approval Level

  • Which translation is most recent? Newer entries often reflect updated terminology or improved style
  • Check workflow role metadata: A segment confirmed by a senior reviewer typically has higher authority than one confirmed by a junior translator
  • Note: recency alone isn't decisive—a recent error is still an error

Translation Quality and Fluency

  • Which translation reads more naturally in the target language?
  • Does one sound like awkward "translationese" while the other is idiomatic?

Step 3: Choose the Master Entry

Based on your analysis, select one translation as the "master" for each conflict group. This is the translation you want the TM to suggest going forward.

Decision criteria, in priority order:

  1. Correctness: The master must be accurate. Eliminate any translation with factual errors or mistranslations first
  2. Alignment with current standards: Choose the translation that matches your current glossary, style guide, and brand voice
  3. Context independence: If possible, choose the translation most likely to work across multiple contexts (unless you're using context-aware matching)
  4. Naturalness and fluency: Prefer idiomatic phrasing over literal translation
  5. Recency: If all else is equal, the most recent translation often reflects the latest thinking

When there's no clear winner: If multiple translations are equally valid for different contexts (e.g., formal vs. informal register), consider splitting them by context metadata or project domain. Alternatively, consult a senior linguist or convene a terminology review meeting to establish a single preferred default.

Step 4: Merge or Delete Non-Master Entries

Once you've chosen the master for each conflict group, decide what to do with the alternatives:

Merge vs. Delete

Merge when:

  • Non-master entries contain valuable metadata (comments, usage notes, custom fields) that you want to preserve
  • You want to retain a record that an alternative translation existed, for future reference

Most TM tools support merging metadata from non-master entries into the master:

  • In memoQ: When filtering for duplicates, choose "Mark for merging" and configure metadata handling (e.g., "Concatenate using commas" to preserve notes from all entries)
  • In Trados: Use import/update policies to specify whether incoming metadata should be merged or overwritten

Delete when:

  • The non-master translation is objectively incorrect
  • It provides no additional value (identical source and target, no unique metadata)
  • Keeping it would create confusion or clutter

Warning: Don't delete entries unless you're certain they're fully redundant. Once deleted, the translation and its metadata are gone. If in doubt, merge instead.

Batch Processing

For large conflict sets (hundreds or thousands of groups), batch operations are more efficient than resolving conflicts one by one:

  1. Export the conflict groups to a spreadsheet (most TM tools support TMX or XLSX export)
  2. Add a column for "Master" or "Action" (merge, delete, keep)
  3. Review and tag entries in bulk (filter by domain, date range, or translator)
  4. Re-import the deduplicated TM using your tool's import function

Step 5: Document the Decision

Resolving the conflict in the TM is only half the work. If the conflict reflects a broader terminology or style choice, document it to prevent recurrence.

Update Your Terminology Database

  • Add the preferred term to your glossary or termbase with a status of "preferred"
  • Mark deprecated alternatives as "forbidden" or "obsolete" with a note explaining why
  • Include usage context if the preferred term is context-dependent (e.g., "Use 'cloud storage' in marketing copy; 'object storage' in technical docs")

Update Style Guides

  • If the conflict revealed a gap in your style guide (e.g., no guidance on formal vs. informal register), add a rule
  • Document edge cases and exceptions so future translators don't recreate the same conflict

Communicate the Decision

  • Notify your translation team of the resolution, especially for high-visibility terms
  • If a popular term was deprecated, explain the reasoning to prevent confusion

Step 6: Prevent Future Conflicts

Conflict resolution is a maintenance task, not a one-time fix. Without preventive measures, conflicts will reaccumulate within months. Build these practices into your workflow:

TM Configuration

  • Disable "allow multiple translations" unless you have a specific use case: This setting permits unlimited alternatives for the same source, which amplifies conflicts. Most teams should leave it off
  • Enable context matching: Store context metadata (preceding/following segments) with each TM entry so the tool can distinguish 100% matches from 101% context matches. This reduces inappropriate reuse across different contexts
  • Disable or carefully manage role-based duplicates: If your TM stores workflow role metadata, configure it to overwrite the existing entry when the same segment is confirmed in a different role (rather than creating a new entry)

Terminology Governance

  • Maintain a single source of truth for terminology: Use a termbase or glossary that's updated before each project and synced with the TM
  • Version your glossaries: Date each glossary revision. When terminology changes, batch-update the TM to replace old terms rather than letting both coexist
  • Flag provisional terms: If a term is under discussion or awaiting client approval, mark it as "provisional" in the glossary and TM. Require review before finalizing

Approval Workflows

  • Require senior linguist approval for new TM entries in high-stakes domains: Restrict write access to the TM. Junior translators can submit suggestions, but only approved entries get stored
  • Implement pre-translation TM review: Before launching a large project, audit the TM for conflicts in the relevant domain. Resolve them proactively rather than during translation
  • Use QA checks during translation: Configure your CAT tool to flag segments where the translator deviates from a high-confidence TM match. Review these segments for potential terminology drift or emerging conflicts

Regular Audits

  • Quarterly light audits: Run automated duplicate/conflict checks. Resolve high-frequency conflicts immediately
  • Annual deep audits: Review a sample of conflict groups manually to catch subtle issues (different registers, evolving style) that automated checks miss

For a broader audit framework, see our guide on auditing translation memory quality, which covers sampling strategies, accuracy checks, and cleanup criteria beyond conflicts alone.

Why Conflicts Keep Coming Back

Even with rigorous conflict resolution and governance, TM conflicts are a recurring problem because the architecture itself is conflict-prone. Translation memory treats each segment as an independent unit, storing outputs without the context or reasoning that made them appropriate.

When the same source segment appears in different contexts—a technical spec vs. a user guide, a formal report vs. marketing copy—a single "correct" translation often doesn't exist. But TM databases aren't built to handle that nuance. They store alternatives and leave the decision to the translator, every time.

This is one reason we built Aquilla around a different model. Instead of matching strings, Aquilla learns from the full context of your project—the surrounding text, your team's documented translation philosophy, and the decisions you've made so far. When it generates a suggestion, it reflects your team's approach to this specific project, not a generic match from a database.

For teams tired of maintaining a conflict-prone TM—especially in long-running, high-stakes projects like Bible translation—this approach reduces the governance burden and improves consistency from the ground up.

Summary: Conflict Resolution Checklist

Use this checklist to guide your conflict resolution workflow:

  • ☐ Identify conflicts using TM filters (sources equal, roles different, etc.)
  • ☐ Prioritize high-frequency segments and key terminology
  • ☐ Analyze each conflict: accuracy, terminology consistency, context, recency, fluency
  • ☐ Choose the master entry based on correctness, current standards, and naturalness
  • ☐ Merge or delete non-master entries (preserve metadata when merging)
  • ☐ Update terminology database and style guides with the decision
  • ☐ Notify translation team of high-visibility resolutions
  • ☐ Configure TM settings: disable multiple translations, enable context matching, manage role duplicates
  • ☐ Establish approval workflows and pre-translation TM review
  • ☐ Schedule quarterly conflict audits

Related Resources