Cooperation before a single tool could do everything
LISA's historical self-description records industry cooperation beginning in 1990. Publishers and language-service organizations exchanged information about production, technology and practical localization procedures. This is an archived organizational account, not a description of an association operating today. W3C's historical LISA introduction.
Imagine a multilingual project moving from one supplier to another. The new supplier needs more than the finished documents. It may need previous sentence pairs, approved terminology and the rules used to split text into translation units. If those resources cannot travel, the team repeats work and loses context.
Interchange specifications make that handover more explicit. Their value is not that every system becomes identical. Their value is that systems can share a defined representation of information.
The technical legacy: several resources, several formats
The LISA OSCAR archive preserves TMX, SRX, TBX, GMX-V and xml:tm materials. Its licensing notice describes the release of the archived portfolio under Creative Commons Attribution 3.0 and identifies ETSI LIS as the designated successor for the standards portfolio. That historical designation does not mean every later project was completed. LISA OSCAR archive.
The distinctions remain useful:
| Resource or concern | Reference to explore |
|---|---|
| Previously translated segment pairs | TMX |
| Concepts and multilingual terminology | TBX |
| Rules for segmenting text | SRX |
| Reporting translation-related volume | GMX-V |
| Document-linked text memory | xml:tm |
These are different technical concerns. A terminology export is not a replacement for a translation memory, and a volume report does not measure whether a sentence was translated correctly.
2011 and the paths that followed
An ETSI proposal records LISA's insolvency effective 28 February 2011 and the initiative to continue standards work in a new setting. The important historical point is a transfer effort, not the disappearance of every specification. ETSI LIS proposal, May 2011.
The subsequent ETSI work register records published TMX and GMX-V deliverables. It also records stopped work items for SRX, TBX and xml:tm. The public record therefore supports a more careful account than “ETSI updated all LISA standards.” ETSI LIS work programme.
TBX has a particularly visible present-day reference: ISO 30042:2019. The TBX community identifies the 2019 format as version 3. ISO lists the edition as published with revision work under way. ISO record, TBX community overview.
Unicode also hosted work on localization interoperability. Its ULI page explicitly says the subcommittee was disbanded in July 2020 and its material remains read-only. An archive that can still be visited is not necessarily an active maintenance organization. Unicode ULI archive.
From a quality model to configurable evaluation
Technical exchange addresses whether information travels. Quality evaluation addresses whether the resulting content meets its intended requirements.
MQM's own history identifies the LISA QA Model as an influence. MQM developed through the QTLaunchPad and QT21 projects, with a modular error typology rather than a single fixed metric. The history also documents DQF-MQM harmonization and its subsequent replacement by MQM Core. MQM Council history.
The conceptual change matters. A museum label, a maintenance instruction and an advertising headline may need different evaluation priorities. An evaluator needs to know the purpose and requirements before deciding how to classify and weight findings.
This does not make quality arbitrary. It makes the assumptions visible. Two scores become meaningfully comparable only when their underlying evaluation settings are sufficiently aligned.
Service processes and project requirements
Not every relevant standard descends from LISA. The broader standards landscape also addresses planning, resources and delivery.
ISO 17100:2015 sets requirements for translation services; its published scope excludes raw machine translation followed by post-editing. ISO 18587:2017 addresses full human post-editing of machine translation. Both have revision activity, so their current published scopes must be distinguished from future proposals. ISO 17100, ISO 18587.
For the planning stage, ISO 11669:2024 gives guidance on translation projects and specifications. For evaluating output, ISO 5060:2024 focuses on human analytic assessment. These are complementary perspectives, not a linear sequence of replacement standards. ISO 11669, ISO 5060.
Localization is more than translated sentences
The OASIS XLIFF committee develops a format for localization interchange. Its work belongs alongside the LISA legacy, but XLIFF should not be attributed to LISA. The committee identifies XLIFF 2.2 as a Committee Specification published in March 2025, while XLIFF 2.1 holds OASIS Standard status. OASIS committee record.
Other references solve adjacent problems. W3C ITS describes localization-related metadata. Unicode CLDR supplies locale conventions, while MessageFormat addresses dynamic messages. Language tags identify the language or variant intended. These layers help explain why a grammatically correct translation can still appear inside an unsuitable product experience. Explore ITS, CLDR, MessageFormat and BCP 47.
A selective timeline
| Period | Milestone | Meaning |
|---|---|---|
| 1990 onward | LISA industry cooperation | Shared production and interoperability problems |
| 2000s | LISA exchange specifications and QA model | Explicit resources and assessment conventions |
| 2011 | Insolvency and continuation initiatives | A branching history, not a complete reset |
| 2012–2013 | Published ETSI GMX-V and TMX documents | Specific continuations of portfolio work |
| 2013 | W3C ITS 2.0 Recommendation | Localization metadata for web content |
| 2015–2017 | ISO service and post-editing references | Requirements for different production processes |
| 2019 | ISO 30042 second edition | A newer TBX reference |
| 2024 | ISO 11669 and ISO 5060 publications | Project specifications and output evaluation |
| 2025–2026 | New specifications and revision projects | Continued work alongside AI-assisted production |
This is a selection of milestones. It is not a claim that every development follows from the previous row.
What AI changes, and what it does not settle
Research such as GEMBA-MQM explores using language models to identify error spans without a human reference translation. The authors also identify limitations associated with a proprietary model. Research results are evidence for a method under particular conditions, not universal assurance. GEMBA-MQM paper.
Our practical interpretation is that AI increases the importance of traceability. An evaluation should identify its inputs, criteria, evaluator configuration and verification method. Otherwise, a score may be impossible to interpret or reproduce.
Existing references still answer different parts of the problem. A file-format specification does not prove that an AI translation is correct. A service standard does not automatically validate every automated workflow. A quality framework does not select an acceptance threshold for every possible use.
The continuity this project follows
We follow continuity of knowledge: exchangeable resources, explicit requirements and inspectable assessment. We are an independent educational project, not a legal successor to the historical language-industry LISA. The history of this domain does not establish such a connection.
The most useful next step is to select the layer relevant to your problem. The standards catalogue provides that route, with separate guides for current references and historical material.
Attribution
The account of MQM's development paraphrases material by the MQM Council, available under CC BY 4.0. It has been abridged and placed in this article's broader historical narrative. See the Council's licensing notice. No endorsement is implied.