CONSULTANCY: Machine Translation

POST
CONSULTANCY: Machine Translation & Workflow Consultant
GRADE: N/A
DUTY STATION: Geneva, Switzerland
COMMENCEMENT OF DUTY: To be determined
NATURE OF APPOINTMENT: Variable (depending on needs of Department/Division)
6 months (50%)

ORGANIZATIONAL UNIT: Linguistic Services and Publication (LSP)
Governance Services Department
WMO is committed to achieving diversity and a balanced workforce. Applications are welcome from qualified women and men, including those with disabilities. The statutory retirement age after 1 January 2014 is 65. For external applicants, only those who are expected to complete the term of appointment will normally be considered.

DUTIES AND RESPONSIBILITIES:
Under the supervision of the Linguistic Support Services Officer, the incumbent will perform the following duties:

A) First part: retraining of Microsoft Custom Translator models:

1. Export WMO bitexts from eLUNa, WMO computer-translation assisted tool, for the years 2022-2024 for the following language pairs: English > Arabic, Chinese, French, Russian and Spanish;
2. Clean up the exported bitexts by removing duplicates, same language matches, empty targets, misalignments, etc.;
3. Retrain Microsoft Translator models per above language pairs (English > Arabic, Chinese, French, Russian and Spanish) and train the reversed models (Arabic, Chinese, French, Russian and Spanish > English);
4. Evaluate the retraining results based on automatic scoring (BLEU/COMET) for the retrained Microsoft models and provide a report on scoring results to LSP;
5. In cooperation with WMO linguists, complete human evaluation of the retrained Microsoft models for the five language pairs listed above;
6. Provide a report on human evaluation of retrained Microsoft models;
7. In cooperation with Linguistic Services Support Officer, prepare bilingual glossaries based on WMO terminology and acronyms for integration into Microsoft Translator models;
8. Retrain all Microsoft Translator models using the prepared glossaries for each language pair;
9. Evaluate the results of retraining based on the integration of glossaries for each language pair and provide a report on retraining outcome;
10. If the retraining results are not satisfactory, take action to improve the existing engines to the extent possible, or conduct research and recommend an alternative solution;
11. Train the Linguistic Services Support Officer and any relevant staff member as deemed necessary by LSP on retraining models and provide guidance material for future retraining and engine maintenance;
12. Present the results of retraining to LSP stakeholders;
13. Document the procedure for retraining and evaluation of models;
14. Carry out other relevant duties as required.

B) Second part: Implementation of a custom translation service in Azure using Power Automate flow-based to connect to the custom Microsoft Custom Translator:

1. Implement a translation workflow in Azure using a Power Automate flow-based to connect to the custom Microsoft Translator machine translation models, allowing LSP staff to translate entire documents (Word docs, Excel spreadsheets, PowerPoint presentations) from a given SharePoint location or pasted text;
2. Create a custom Microsoft Form for translation of pasted text or launching the workflow for translation of entire document(s) based on given criteria (selection of languages, selection of folder, etc.);
3. The form should give the user the option of selecting whether to use standard Microsoft Translator engine and/or trained custom Microsoft Translator models;
4. Translated documents should be saved in a given SharePoint location;
5. Translated documents should be marked by a disclaimer (the text to be provided by LSP) in the target language;
6. Notification of completion of task or error to be sent to specified email address(es);
7. Make the necessary tests and adjustments to ensure proper functioning of the flow and Azure interface for all the language pairs;
8. Train the Linguistic Services Support Officer and any relevant staff member as deemed necessary by LSP on Azure and Power Automate flow and other related components;
9. Organize a training workshop for the stakeholders involved in the process to give an overview of the project;
10. Document the implemented procedures and workflows and other related components adapted to LSP needs;
11. Carry out other relevant duties as required.

QUALIFICATIONS:
Education: University degree in information technology, computer science, language technology or related field.
Experience: Proven experience in:

1. Training and retraining of machine translation models (Microsoft and other engines).
2. Developing a custom translation workflow and front-end form utilizing Microsoft tools: Azure, Power Automate, custom Translator, SharePoint, Forms, and other Microsoft resources, as necessary.
3. The contractor should have at least 2 or 3 years of experience.

Other requirements:

1. Knowledge of training and retraining of machine translation models (Microsoft Translator required, other engines a distinct advantage).
2. Knowledge of Microsoft tools (Azure, Sharepoint, Forms, Power Automate, etc.).
3. Ability to set up flows with Power Automate.
4. Ability to transfer knowledge by documenting processes and providing training.

Languages:
Excellent knowledge of English (both oral and written). Knowledge of other official languages of WMO would be an advantage.
(Note: The official languages of the Organization are Arabic, Chinese, English, French, Russian and Spanish.)

Additional Information:
Focal point email for additional information: seymann@wmo.int
Applications should be made online through the WMO e-recruitment system at https://erecruit.wmo.int/public. Do not send your application via multiple routes. Only applicants in whom WMO has a further interest will be contacted. Shortlisted candidates may be required to sit a written test and/or an interview.

#J-18808-Ljbffr

---