Senior Research and Development Engineer
I joined Idiap Research Institute in February 2008 as a Development Engineer. Before that I had the opportunity to work in three startups in the United Kingdom, also as a Development Engineer.
My first development tasks at Idiap were some low level server socket java programming on a European Project called AMI.
Then my work shifted to Optical Character Recognition integration and customization with different commercial partners.
In 2012 I started to specialize into Automatic Speech Recognition. During this time, with the help of Idiap researchers, we built a bilingual Automatic Speech Recognition system for the Valaisan Parliament in order to automatically transcribe and index political debates.
Between 2013 and 2016 my activities were essentially related to Speech Technologies, i.e. voice activity detection, acoustic modeling training with HTK, Kaldi, Quicknet toolkits (GMM, SAT, Discriminative training, ANN and DNN), acoustic modeling adaptation (MLLR, MAP), language detection (French, German) and language modeling using various toolkits (IRSTLM, MITLM, SRLM).
I also spent a considerable amount of time trying to make the output of ASR systems usable by humans. For example, we designed frameworks to automatically format speech recognition results (numbers, dates, acronyms, ...) and predict punctuations marks like dots into ASR output using grammar and silences information.
In 2017 my work was focused on Applied Research in Automatic Speech Recognition for Swiss German dialekts. I also trained an I-Vector system used in a Speaker Diarization pipeline.
I am currently doing some Applied Research in Punctuation Prediction in ASR transcripts using Deep Learning.
In 2014 my title changed to Senior Development Engineer.
In 2018 my title changed to Senior Research and Development Engineer.
For additional information, please see my Linkedin profile.
I have strong interests in Deep Learning, Machine Learning, Data Science, Statistical Inference, Data Modeling and Uncertainty Strategies i.e. complex systems, agile software development and agile project management.
I am in a possession of a bachelor degree (HES) in Computer sciences and Business from the University of Applied Sciences in Sierre.
Empirical Evaluation and Combination of Punctuation Prediction Models Applied to Broadcast News
Authors: Alexandre Nanchen; Philip N. Garner
Comparative Study on Sentence Boundary Prediction for German and English Broadcast News
Authors: Wang Yang, Nanchen Alexandre, Lazaridis Alexandros, Imseng David, Garner Philip N.
Automatic Accentedness Evaluation of Non-Native Speech Using Phonetic and Sub-Phonetic Posterior Probabilities
Authors: Ramya Rasipuram, Milos Cernak, Alexandre Nanchen, Mathew Magimai Doss
Exploiting un-transcribed foreign data for speech recognition in well-resourced languages
Authors: David Imseng, Blaise Potard, Petr Motlicek, Alexandre Nanchen, Herve Bourlard
Automatic Speech Indexing System of Bilingual Video Parliament Interventions
Authors: Gyorgy Szaszak, Milos Cernak, Petr Motlicek, Alexandre Nanchen, Philip N. Garner, Flavio Tarsetti
MediaParl: Bilingual mixed language accented speech database
Authors: Imseng David, Bourlard Herve, Caesar Holger, Garner Philip N., Lecorve Gwenole, Nanchen Alexandre
A Speech-based Just-in-Time Retrieval System using Semantic Search
Authors: Popescu-Belis Andrei, Yazdani Majid, Nanchen, Alexandre, Garner Philip N.
Open source software
During this years I had the opportunity to build tools to facilitate the preparation of data required to train ASR based systems. Two of them are open source: ASRT (Automatic Speech Recognition Tools) and Webvalidation (annotation tool).