midobal/mthd

A demo showcasing two neural machine translation applications to historical documents.

Hi there, welcome to my GitHub profile! I’m Miguel Domingo. I work as a Ph.D. researcher at the PRHLT Research Center from Universitat Politècnica de València. I have been mainly working in machine translation and historical document processing, but my interests also include other areas of machine learning, computational linguistics and pattern recognition. You can find more information about me at my website.

---

Here you can find highlighted what I consider to be some of my most useful contributions.

Research

• Dockerfiles: a collection of custom Dockerfiles for several tasks.
• Interactive Neural Machine Translation: a fork of the OpenNMT-py toolkit implementing several interactive machine translation protocols.
• MT Scripts: a collection of several scripts useful for research in machine translation.
• Statistical Dictionary: Given a parallel data set, this software trains a statistical dictionary for translating documents. It was built in Python and makes use of mgiza for computing IBM model 1 alignments.
• Word-Level AutoCompletion: a fork of the OpenNMT-py toolkit that deploys a word-level autocompletion tool.

Teaching

• Neural Machine Translation: a Practical Session: a Google Colab notebook which presents a practical session to introduce the training of neural machine translation systems to students.
• Stable Diffusion Fine-tuning with DreamBooth: a Google Colab notebook which presents a practical session to fine-tune Stable Diffusion so it learns to generate yourself.

Templates

• Dissertation: the LaTeX template I designed for writing my Ph.D. dissertation.
• Jekyll Theme: custom tweaks to Jekyll's minima theme based on academic portfolio.
• TwoCol Style: a custom two-columned LaTeX style template based on the ACL Style template.