ML4HMT

Shared Task on Applying Machine Learning Techniques to
Optimise the Division of Labour in Hybrid MT (ML4HMT-2011)

Barcelona (Spain) · Saturday, November 19th, 2011

In conjunction with International Workshop on Using Linguistic Information for Hybrid Machine Translation (LIHMT-2011)

Shared Task Purpose

The "Shared Task on Applying Machine Learning Techniques to Optimise the Division of Labour in Hybrid MT" is an effort to trigger systematic investigation on improving state-of-the-art Hybrid MT, using advanced machine-learning (ML) methodologies. The main focus of the shared task is trying to answer the following question: Can Hybrid/System Combination MT techniques benefit from extra information (linguistically motivated, decoding and runtime) from the different systems involved?

Workshop Details

Location
Barcelona Media
Av. Diagonal, 177
08018 Barcelona, Spain

Organisational

Registration
Participation is free!

Programme

You can download the workshop programme (45 KB) here.

9:15Welcome
9:30Toni Badia (BM) — Introduction to the ML4HMT Shared Task Workshop
9:40Patrik Lambert (LIUM) — The MANY System @ML4HMT-2011 (200 KB) · presentation slides (786 KB)
10:30Tsuyoshi Okita (DCU) — DCU System Combination @ML4HMT-2011 (196 KB) · presentation slides (209 KB)
11:00Eleftherios Avramidis (DFKI) — DFKI System Combination with Sentence Ranking @ML4HMT-2011 (108 KB) · presentation slides (614 KB)
11:30Coffee Break
12:00Christian Federmann (DFKI) — DFKI System Combination using Syntactic Information @ML4HMT-2011 (152 KB) · presentation slides (748 KB)
12:30Christian Federmann (DFKI) — Comparison of Overall Results @ML4HMT-2011 (532 KB) · presentation slides (1 MB)
12:40Alon Lavie (CMU) — MEMT: Alignment-based MT System Combination with Linguistic and Statistical Features · presentation slides (1.1 MB)
13:10

Discussion Panel Chair: Patrik Lambert (LIUM), Alon Lavie (CMU), Cristina España-Bonet (UPC) and Christian Federmann (DFKI). Topics include:

  1. Two Hybrid paradigms: Multi- vs Single-system
  2. In the Multi-system approach: can Hybrid/System Combination MT techniques benefit from extra information (linguistically motivated, decoding and runtime) from the different systems involved?
  3. Evaluation in the Multi-system approach: do we evaluate the output in isolation or do we use evaluation information from the different systems involved?
14:00Lunch