Institute for Translation and Interpreting, Heidelberg University
Summer Semester 2020-2021
Course 0945MUE08211 Projektarbeit: Forschung: Evaluation of Machine Translation systems in task-critical domains
- 0945MUE08211 SoSe 2021 Projektarbeit Forschung: Evaluation of Machine Translation systems in task-critical domains; MA, 2nd Semester, 6CP, Main seminar, Mon 14:15 to 15:45
- LSF link
Submitting your assessments:
Area on Moodle to submit your assessments: https://moodle.uni-heidelberg.de/mod/assign/view.php?id=447062
If you do not have the Heidelberg UniID (being from another institution), please submit your essay via this link: https://heibox.uni-heidelberg.de/u/d/ade488d70a8448ffa40f/
- Deadline 16.08.2021, 23:59 AOE time
- (if any problems with Moodle, please send your assessment to me via email)
Course enrolment on Moodle
0945MUE08211
Virtual Seminar Room
We start this course on-line. Please join our Seminar Room:
https://heiconf.uni-heidelberg.de/bab-eeq-ax3
on Mondays, 14:15 Central European Summer Time. If any problems with computer audio, you can also dial in via the phone: +49 6221 431 0 410, and then enter Phone PIN: 4341-9332-8165
Course materials
Week01 - Introduction:
Week02:
Week03:
Week04:
Week05:
- Video recording of the class: https://youtu.be/PxQB14_bFxE
- On Monday, 10.05.2021, we will start as usual, 14:15, but we will have to finish earlier, at 15:00. To make up for this time, I invite you to watch two talks relevant for our topic (datasets and workflows with the use of machine translation):
- “Language Industry Transformation: Now and Tomorrow”, by Translation Automation User Society (TAUS) team: https://www.youtube.com/watch?v=ByUJWaPcha4&list=PL6uIqKfoBxnh7iki2JMuzENNXLaeekB2t&index=11
- “API-driven translation workflows”, by Daniel Gray, Daniel Italiaander, One Hour Translation: https://www.youtube.com/watch?v=CZcwgaulnWw&list=PL6uIqKfoBxnh7iki2JMuzENNXLaeekB2t&index=9
Week06:
Week07: Public holiday, no class
Week08:
Week09:
- This class will be off-line, there is a take-home practical task instead of the class, which has to be completed by Sunday 11.06.2021 (next week).
- Before doing the task, you can watch the video on YouTube for this week: https://youtu.be/tpBNc-1ALkA. If you missed last week, you can catch up by watching video under Week08
- Then download and fill in the worksheet Word file from last week https://heibox.uni-heidelberg.de/d/6aeef8df547d4c78ad6e/files/?p=%2F0945MUE08211-y20s21w08-worksheet-research-project-v02.docx&dl=1. ( All the instructions are in the HeiBox folder shared under Week08 link).
- You do not have to pre-edit all 47 segments, but please try to pre-edit and evaluate usability for post-editing of at least 12, or as many as you can.
- Use this Google Form https://forms.gle/Y9uf6aZmmYjkCc6eA 4 times, to evaluate:
- the baseline text without pre-editing (e.g., translated with Bing Translator https://www.bing.com/translator or another MT system without controlled language)
- the text pre-edited according to CL rules
- the baseline text translated by another (e.g., with Google MT, DeepL, etc.)
- the ‘surprise system’: CL-pre-edited text, translated by another MT system (each time fill in a new ‘response’ into the same form)
- When finished, re-name your worksheet file, using your initials or name and your target language
- Upload your worksheet to the HeiBox link: https://heibox.uni-heidelberg.de/u/d/5dff43c513e64d499706/
- Go to the download link https://heibox.uni-heidelberg.de/d/509eccffe18b4a39a6c3/ and download your colleague’s or your tutor’s worksheet, which will contain CL developed for another target language.
- Repeat your evaluation experiment using the text in the CL column by filling the same Google form
- You can check the evaluation results given by you and your colleagues on Google Spreadsheet (read only) link https://docs.google.com/spreadsheets/d/1hC9ZH-4B5I3JQ2e2E2ykiuXqif23dYMrhxDnlepEMWE/edit?usp=sharing. You can download and calculate average, etc. scores in the spreadsheet.
- Once again, please upload your Worksheet Word file and your evaluation scores on Google Form by Sunday 11.06.2021 (next week)
Week10:
Week11:
Week12:
Week13:
Week14:
- no class, the week for preparation of the MT evaluation / error analysis project
Week15: