|
| 1 | +--- |
| 2 | +# Documentation: https://wowchemy.com/docs/managing-content/ |
| 3 | + |
| 4 | +title: "Seminar: \"Prudent NLG Evaluation with Humans\"" |
| 5 | +# event: |
| 6 | +# event_url: |
| 7 | +location: Abacws |
| 8 | +# address: |
| 9 | +# street: |
| 10 | +# city: |
| 11 | +# region: |
| 12 | +# postcode: |
| 13 | +# country: |
| 14 | +summary: Talk by [Vilém Zouhar](https://vilda.net/) (ETH Zürich, Switzerland) |
| 15 | +abstract: "Annually, research teams spend large amounts of money to evaluate the quality of NLG systems (WMT for machine translation, inter alia). We'll first look at how to speed up and improve the quality of the annotators' work by pre-filling annotations with automatic quality estimation ([ESA](https://aclanthology.org/2024.wmt-1.131/), [ESAᴬᴵ](https://arxiv.org/abs/2406.12419)). In the second part, we'll take the automatization a step further and try to determine which segments do not need to be evaluated at all. For this, we make use of methods from psychometrics for efficient yet informative testset construction for human students. In our case, the students to be tested are NLG systems." |
| 16 | + |
| 17 | +# Talk start and end times. |
| 18 | +# End time can optionally be hidden by prefixing the line with `#`. |
| 19 | +date: 2025-01-16T13:00:00Z |
| 20 | +date_end: 2025-01-16T14:00:00Z |
| 21 | +all_day: false |
| 22 | + |
| 23 | +# Schedule page publish date (NOT event date). |
| 24 | +publishDate: 2025-01-13T00:00:00Z |
| 25 | + |
| 26 | +authors: [alvamanchegof] |
| 27 | +tags: [] |
| 28 | + |
| 29 | +# Is this a featured event? (true/false) |
| 30 | +featured: false |
| 31 | + |
| 32 | +# Featured image |
| 33 | +# To use, add an image named `featured.jpg/png` to your page's folder. |
| 34 | +# Focal points: Smart, Center, TopLeft, Top, TopRight, Left, Right, BottomLeft, Bottom, BottomRight. |
| 35 | +image: |
| 36 | + caption: "" |
| 37 | + focal_point: "" |
| 38 | + preview_only: false |
| 39 | + |
| 40 | +# Custom links (optional). |
| 41 | +# Uncomment and edit lines below to show custom links. |
| 42 | +# links: |
| 43 | +# - name: Follow |
| 44 | +# url: https://twitter.com |
| 45 | +# icon_pack: fab |
| 46 | +# icon: twitter |
| 47 | + |
| 48 | +# Optional filename of your slides within your event's folder or a URL. |
| 49 | +url_slides: |
| 50 | + |
| 51 | +url_code: |
| 52 | +url_pdf: |
| 53 | +url_video: |
| 54 | + |
| 55 | +# Markdown Slides (optional). |
| 56 | +# Associate this event with Markdown slides. |
| 57 | +# Simply enter your slide deck's filename without extension. |
| 58 | +# E.g. `slides = "example-slides"` references `content/slides/example-slides.md`. |
| 59 | +# Otherwise, set `slides = ""`. |
| 60 | +slides: "" |
| 61 | + |
| 62 | +# Projects (optional). |
| 63 | +# Associate this post with one or more of your projects. |
| 64 | +# Simply enter your project's folder or file name without extension. |
| 65 | +# E.g. `projects = ["internal-project"]` references `content/project/deep-learning/index.md`. |
| 66 | +# Otherwise, set `projects = []`. |
| 67 | +projects: [] |
| 68 | +--- |
| 69 | + |
| 70 | +**Invited Speaker:** [Vilém Zouhar](https://vilda.net/) (ETH Zürich, Switzerland) |
| 71 | + |
| 72 | +**Bio:** |
| 73 | +Vilém is a PhD student at ETH Zürich working on both human and automatic evaluation of MT/NLG systems, balancing costs, quality, and bias. |
0 commit comments