My Picture Marco Turchi
Research Interests
Research Interests


Marco turchi

My Research

My current research is centered on sequence-to-sequence neural methods for Neural Machine Translation applied to develop innovative solutions for the  translation industry. My recent works involve the development of neural machine translation and automatic post-editing systems that are able to continuously learning from human feedback.

Before embracing the MT world, I focused my research on  information extraction, retrieval and organization from text documents, using several machine learning (probabilistic and statistical) approaches. I also analyzed different aspects of text/web mining moving from classical text problems as supervised and unsupervised learning, or new representation for text document to more computational linguistic tasks, as statistical machine translation and document summarization.


Neural Machine Translation (NMT)

  • Improving Input Representation.
  • Continuous Learning for NMT.
  • Multi-lingual NMT.
  • Contextual Handling for Text Translation.
  • Automatic Post-Editing.
  • ...

Statistical Machine Translation (SMT)

  • Improving Translation Quality in News domain.
  • Learning Curves of SMT systems.
  • Confidence Estimation.
  • Parallel Sentences Extraction from the Web.
  • Self-Learning for SMT systems.
  • Hybrid system.
  • ...

Natural Language Processing and Text/Web Mining:

  • Feature selection.
  • Text classification.
  • Integration of semantic in text classification.
  • Detection of content in text documents.
  • Clustering techniques.
  • Semi-supervised clustering algorithms.
  • Soft-clustering.
  • Active learning algorithms for text classification. 
  • Time series analysis of textual data.
  • Detection of text pattern into text documents.
  • Extraction and analysis of multilingual text content.
  • Document Summarization.
  • ...

Machine Learning:

  • Smoothing techniques for parameter estimation.
  • Regression and classification algorithms.
  • Learning curves.
  • Kernel methods - String Kernels.
  • Graphical Models.
  • Robust approaches for outlier detection.
  • ...

Language evolution.

High Performance Computing (HPC), large-scale dataset.

Some links:

Svm - Support Vector Machine

- Kernel Methods for Pattern Analysis (here)

- PLS - Partial Least Square

- Graphical Model (here)

- LDA - Latent Dirichlet Allocation

- String Kernels (here)

NLP world:

- ACL - Association for Computational Linguistics

- SMT - Statistical Machine Translation

- Machine Translation Archive (here)

- EAMT - European Association for Machine Translation

Document made with KompoZer