The Speech Technology Group (Grupo de Tecnología del Habla, GTH) is part of the Department of Electronic Engineering (IEL), which belongs to Universidad Politécnica de Madrid (UPM), at the High Technical School of Telecommunication Engineering (ETSIT). UPM has the 'HR Excellence in Research Award' within the European Commission Human Resources Strategy for Researchers (HRS4R) as institution that aligns their human resources policies to the 40 principles of the European Charter & Code for recruitment of researchers. Besides, our group belongs to the IPTC (Information Processing and Telecommunication Center) and AIR4S Innovation Hub for Artificial Intelligence and Robotics.
Our research activities started in 1978, being our group devoted to research and development in various areas of speech science and technology, including speech synthesis and recognition, speaker diarization and identification, language recognition, machine translation, conversational systems, and different aids for the handicapped. Besides, our group also carries out research on combining inertial sensor information and images for detecting different health conditions, on image and video processing for improving tourist experience, or for detecting emotional states on TV shows and social media.
Finally, we are also working on natural language processing applications for reasoning and understanding over text documents (Q&A, paraphrasing, classification), as well as on conversational systems.
*2. Project description and objectives*
Currently we are offering one PhD position in any of these three main research areas:
· *Conversational agents* is an important and *very active research* field that has experienced sustained growth recently thanks to the advances on deep neural network topologies that can handle different aspects of the dialog and system, e.g. contextual memory, reasoning over facts, generation of diverse and emotive answers, and the possibility of adapting the dialog to a given persona profile. However, the integration of all these components is still not perfect creating frustrations on final users and limitations for companies. In this research line, we will study and propose new *robust *mechanisms to make *generative-based social chatbots* to adapt themselves to *new vocabularies and profiles*, while providing answers based on *background information*. Participation in international challenges such as DSTC or ConvAI will be conducted.
· *Reasoning over texts* is an important and very active field that has experienced sustained growth recently thanks to the development of *semantic and contextual* *vector embeddings*, which combined with advanced deep neural network topologies, are used to answer questions given supporting paragraphs, to generate new questions (paraphrasing) or summaries. Besides, these vector embeddings play an important role in other areas like machine translation, conversational systems, speech recognition or language modeling, etc. In this research line, we will study new mechanisms to allow the *incorporation of human knowledge, common-reasoning and priming* over vector embeddings for different kind of documents, sentences and domains. Participation in international challenges such as ConvAI or SemEval will be conducted.
· *Language recognition* is the process of detecting the spoken language or dialects over speech recordings. State-of-the-art systems are based on *extracting acoustic vectors* that only take into account acoustic information from the audio signal for detecting language without considering higher-level information as syntactic information, phoneme sequence information, etc. So, *phonotactic information* from the sequence of phonemes in a sentence can be used to complement the results obtained with the acoustic models. In this project, we will propose *new mechanisms to improve the acoustic vectors isolated from channel and noise effects*, while applying *advanced NLP techniques over the phonotactic information* and test it over well-known datasets used on international competitions (NIST or Iberspeech).
All these project goals will allow the candidate to move beyond just research into practical systems that can be used in a wide range of text processing applications, creating a *huge impact* for the *scientific* (top journals and conference papers) and *industry* (patents, licenses) communities.
*3. What do we offer?*
- A four-year fully funded PhD position with yearly progress evaluation,
- A competitive salary (including holiday allowance) and benefits in
accordance with qualifications and experience (approximately 1600
Euro/month gross amount),
- An international scientific environment driven by excellence in
- Opportunities for travelling to conferences and research visits to
international partner research groups.
*4. Candidate requirements*
Candidates are required to:
1. Hold a master’s degree in computer science, computer or telecommunication engineering, physics, mathematics or equivalent,
2. An excellent academic record,
3. Proven programming experience (e.g. Python, C++),
4. Strong teamwork skills,
5. Candidates should have a fluent competency in written and spoken English. Knowledge of Spanish is not compulsory--but the candidate’s competency or willingness to learn this language will be considered an asset.
We will consider as plus the following characteristics:
1. Experience with deep learning frameworks (pyTorch preferably),
2. Demonstrate prior publishing experience, preferably with at least one paper at one of the top-tier venues in NLP, Speech, or machine learning (ACL, EMNLP, NeurIPS, ICASSP, Interspeech, ASRU) or other specialized conferences (SigDial, LREC, IWSDS),
3. Experience in taking part in challenges such as NIST LRE or SRE, Iberspeech, ConvAI, AlexaPrize, DSTC, DBDC, etc.
*5. How to apply*
Interested candidates should provide the following documentation:
1. A detailed CV (with complete contact details),
2. A motivation letter related to the offered projects and how the candidate will contribute to it (max 2 pages),
3. An academic record from BSc and MSc studies,
4. Complete contact details (name, title, mail and telephone) for two referees.
All documentation should be submitted via email to Luis Fernando D’Haro ( luisfernando.dharo at upm.es <agonzalez at unav.es>) and Ricardo Córdoba ( ricardo.cordoba at upm.es).
*Submission deadline:* April 31, 2020 or until the position is filled
Luis Fernando D’Haro (Associate Professor)
Phone: +34 9106 72174
Twitter: @lfdharo -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: text/html Size: 29201 bytes Desc: not available URL: <https://mailman.uib.no/public/corpora/attachments/20200302/910300b6/attachment.txt>