The CoML team is currently recruiting post-docs, doctoral students, master’s and engineering school internships to work on the topics described above. Inside those topics, short internships (4-6 months minimum), master internships or PHD projects can be defined. Some of the projects can be conducted in collaboration with our colleagues in Tokyo (Riken Institute, University of Keo), at Johns Hopkins University (USA), at Brno University (Czech Republic), or, for the more machine learning savy, Facebook AI Research, Google or Microsoft.
Developing machine learning, computational linguistics or signal processing techniques to large corpora of spontaneous speech in typologically diverse languages in order to model the unsupervized discovery of linguistic structure in the human infant. Possible topics (not exhaustive):
modeling neural processing in the auditory cortex, context effects, inverse articulation modeling
unsupervized discovery of spoken terms, of phonetic, prosodic, semantic or grammatical categories
modeling the effect of external feedback loops (babbling, social, interactions)
testing the universality of learning algorithms across very different languages and the impact of multilingual input
Background required: engineering, computer science, applied maths, physics or formal/computational linguistics. Familiarity with cognitive/language science is a plus.
Applying machine learning tools to analyze ‘ecological’ datasets (daylong recordings of casual, interactive speech), to make predictive models of language acquisition (in young children) or language degradation (in patients with neurodegenerative pathologies). Test these predictions against behavioral results (eye tracking, vocabulary size) or brain imaging data (EEGs, fNIRS).
Background required: neurosciences, experimental psychology or medicine. A good command of programming and statistical data analysis is clearly a plus.
Developing ‘psychology’ tests comparing human learning and machine learning in a variety of language and cognitive tasks (phonetic, lexical, syntactic, semantic, pragmatic processing, common sense reasoning). The idea is to develop a human vs machine benchmark on key areas where humans display superior flexibility and adaptability.
Background required: developmental psychology, linguistics, cognitive psychology. Some projects require design skills. All a good command of programming.
A limited number of positions for a Master’s internship are open every year. These internships should be for a minimum of 5-6 months. (for Topic 1, see here; for other Topics, contact us).
If you are interested in applying to a relevant master, here is a non limitative list): Cogmaster, Dual Master, BIP Neuroscience, MVA, etc. For international students, there is the possibility of funding through the ENS/IEC program or the ENP program.
A limited number of positions are open for engineering internships throughout the year (last year internships, master’s internships, etc.). These internships should be for a minimum of 4 months, preferably for 5-6 months. They concern Topic 1 (see detailed announcement here).
The project is currently recruiting 2 post-docs for a two year position, with two different profiles.
The first profile is in computational modeling/machine learning, and concerns a joint project between Ecole Normale Supérieure/INRIA in Paris and Microsoft Research (Redmond). The general topic is unsupervised learning of speech representations using several machine learning techniques.
The second profile is in EEG data collection and analysis. It consists in establishing a robust method to assess language development landmarks in infants at the phonological and lexical level. This method will be used to test the empirical predictions made by machine learning algorithms developped in the team and elsewhere.
One doctoral position is open per year (pending competitive funding). It is strongly advised to first do an internship in the lab (Master’s or Engineering) before applying to a doctoral position. Indeed, French PhDs are only 3 year long, and the requirements are high (original research topic, 3 published papers, etc). A preliminary internship enables to explore a topic, gather initial results, and writing a strong thesis proposal.
Regarding funding (3 year fellowships), students will typically apply through the Ecole Doctorale 3C (preselection Feb, selection July), the Ecole Doctorale Frontiere du Vivant (preselection Feb, selection July). Additional possibility for funding are to be found through specialized agencies (eg; DGA) or through an industrial cooperation (CIFRE). The ITI PSL program is a 4 year program which includes a year with courses and internships before the student can apply to a 3 years PHD fellowship.
There is only a limited number of open positions for doctoral students every year. Potential applicants should contact syntheticlearner at gmail dot com for more information.