MECILDI V2
MECILDI V2
The development of the complete computational framework MECILDI V2 have started in July 2026. It will build on MECILDI V1 extending its capabilities to identify as much as possible web technics for designing multilingual websites and follow IETF DCP47 (RFC4447 and RFC5646) in order to process the full range of languages online. It is a complex development which will last several months and plan to conclude by the end of 2026.
The project is only possible thanks to the support of Organisation de la Francophonie (OIF), with additional support of DGLFLF.
The first pilot application of MECILDI V2 will be dedicated to the linguistic measurement and follow-up of 10 ccTLDs covering regional diversity of countries members of OIF, in alphabetic order : Burundi, Democratic Republic of Congo, French Polynesia, Haiti, Ivory Coast, Lebanon, Madagascar, Morocco, Romania and Vietnam
The first step is to measure language repartition and multilingualism of those countries’ ccTLD using MECILDI V1.
The second step is to draw lessons from the MECILDI V1 experience, both in terms of results and implications for V2.
The transition phase between V1 and undergoing V2 has been nurtured by the application of V1 in some 50 different ccTLDs and gTLDs. The lessons learned have also lead us to keep improving V1 and finally to reach V1.9.4 with the following improvements or strategic decisions:
- The experience has shown that many websites appear repeatedly using the redirection technics from various original websites. Those redirected websites could be of several types: domain registries, web hosting in generic platform such as sites.google,com, facebook or instagram, redirect to wikipedia article, global services such as bittly.com or wix.com. some companies also use the redirect technique to give access from different domain names to their web services. The frequent common characteristic of those duplicated redirections is to often have a high level of multilingualism. So we have taken the strategic decision not to allow more than one redirection to the same domain, in order to avoid artificial increase of AvgL. This obviously have impacted the results by decreasing the AvgL in proportion varying depending of the analyzed ccTLD.
- Analyzing many cases of failures to offer meaningful inputs to the language detection module, conducing to the NON-ID classification, some new techniques has been developed to reduce drastically those cases and, with this improvement, we have managed to obtain a NON-ID category close to 0, e.g. full treatment of the cases.
- Analyzing the cases of errors in language detection (like taking Spanish for Guarani because of the presence of some street names) we have also mange some improvements to reduce those errors and turning them from unfrequent to marginal.
The result is a last version of MECILDI V1 with a higher level of accuracy and reliability. We have proceed to a new run of MECILDI on TRANCO with input from October, 20th 2026 and rerun the previous one of June 2026 with the V1.9.4 in order to allow fair comparison and trend analysis. We will progressively replace all published data with their equivalent obtained by V1.9 in order to allow fair comparisons with future run.
October 2026 results for Tranco and comparison with June 2026.
| INDICATORS | RUN13 | RUN12 | CHANGE |
| Pages@English | 20.59% | 20.91% | -1.51% |
| MultiR | 3.29 | 3.35 | -1.61% |
| %Multi | 37.44% | 38.32% | -2.29% |
| AvgL | 7.14 | 7.15 | -0.08% |
| ERR | 44.68% | 47.55% | -6.04% |
| ErrG1 | 13.41% | 15.04% | -10.82% |
| ErrG2-5-6 | 16.06% | 17.68% | -9.14% |
| ErrG3 | 1.91% | 2.12% | -9.83% |
| ErrG4 | 5.24% | 4.75% | 10.26% |
| ErrG7 | 8.05% | 7.96% | 1.16% |
| MonoL | 62.56% | 61.68% | 1.42% |
| GooT | 1.09% | 0.96% | 13.55% |
| IDN | 0.16% | 0.16% | 3.90% |
| RUN13 | %PAGES | %SITES | RUN12 | %PAGES | %SITES | CHANGE | |
| TOTAL | 100.00% | 328.92% | TOTAL | 100.00% | 335.42% | -1.94% | |
| English | 20.59% | 66.85% | English | 20.91% | 68.79% | -1.51% | |
| German | 6.67% | 21.79% | German | 6.58% | 21.85% | 1.26% | |
| French | 6.42% | 20.95% | French | 6.23% | 20.67% | 3.12% | |
| Spanish | 6.33% | 20.70% | Spanish | 6.23% | 20.62% | 1.62% | |
| Italian | 4.16% | 13.59% | Italian | 4.08% | 13.62% | 1.98% | |
| Portug. | 3.89% | 12.75% | Portug. | 3.93% | 13.10% | -0.88% | |
| Russian | 3.56% | 11.64% | Russian | 3.55% | 11.80% | 0.29% | |
| Dutch | 3.28% | 10.75% | Dutch | 3.23% | 10.82% | 1.64% | |
| Chinese | 2.95% | 9.70% | Japanese | 3.01% | 10.03% | -3.54% | |
| Japanese | 2.90% | 9.49% | Chinese | 2.86% | 9.54% | 3.16% | |
| Polish | 2.69% | 8.89% | Polish | 2.63% | 8.81% | 2.30% | |
| Indones. | 1.88% | 6.19% | Indones. | 1.92% | 6.42% | -1.98% | |
| Korean | 1.81% | 5.98% | Turkish | 1.84% | 6.18% | -2.53% | |
| Turkish | 1.79% | 5.89% | Swedish | 1.79% | 6.02% | -0.51% | |
| Swedish | 1.78% | 5.87% | Korean | 1.76% | 5.96% | 2.55% | |
| Arabic | 1.71% | 5.64% | Arabic | 1.76% | 5.97% | -2.98% | |
| Czech | 1.56% | 5.18% | Czech | 1.56% | 5.25% | 0.10% | |
| Danish | 1.40% | 4.61% | Danish | 1.38% | 4.67% | 1.11% | |
| Finnish | 1.35% | 4.45% | Finnish | 1.30% | 4.41% | 3.74% | |
| Romanian | 1.28% | 4.24% | Ukrainian | 1.29% | 4.35% | -1.53% | |
| Ukrainian | 1.27% | 4.23% | Romanian | 1.28% | 4.31% | 0.22% | |
| Hungar. | 1.26% | 4.18% | Hungar. | 1.27% | 4.30% | -1.08% | |
| Vietnam. | 1.20% | 3.98% | Vietnam. | 1.24% | 4.20% | -3.55% | |
| Greek | 1.18% | 3.94% | Greek | 1.18% | 4.00% | 0.11% | |
| Thai | 1.05% | 3.47% | Thai | 1.09% | 3.71% | -3.93% | |
| Slovak | 0.98% | 3.28% | Slovak | 0.99% | 3.35% | -0.09% | |
| OTHERL | 0.98% | 3.32% | OTHERL | 0.94% | 3.31% | 4.92% |
Projects by OBDILCI
- Indicators for the Presence of Languages and multilingualism in the Internet
- The Languages of France in the Internet
- French in the Internet
- Portuguese in the Internet
- Spanish in the Internet
- Web Multilingualism reports
- Courses
- AI and Multilingualism
- Linguistic gTLDs
- DILINET
- Pre-historic Projects…
- Digital Language Death