The MultiSem project will propose novel advanced models for multilingual semantic processing. Existing data-driven models employ robust machine learning techniques for handling vast amounts of textual data but overlook the intricacies of the mechanisms involved in language processing which should be reflected in automatic methods. At the same time, findings in the computational semantics field fail to make their way to large-scale NLP systems, mainly due to the focus on small lexical samples which restricts the potential of the models to scale up and be used on unrestricted text. Interaction between disciplines has thus been limited up to now and the mutual potential benefits of their respective research remain unclear. At this moment of burgeoning interest in multilingual processing and semantics-related research, the MultiSem project proposes to bridge the gap between disciplines by combining the efficiency and robustness of state of the art approaches to semantic analysis with linguistically motivated semantic representations. The main novelty of the semantic processing models proposed in MultiSem is that they will be able to adapt processing to different lexical items and text types, inspired by findings regarding the organisation of semantic information in the mental lexicon and the role of context in meaning activation. It has been shown that instead of considering all possible interpretations for words in context, human bilinguals and translators restrict their choice to specific senses. This focus is largely influenced by the parameters of the communicative context and by the domain and topic of the processed texts, while a finer-grained filtering occurs only when needed for improving text understanding. Based on these findings, the models developed in MultiSem will differentiate semantic processing according to the disambiguation needs of specific words, contexts and textual genres. To achieve this ambitious goal, we intend to combine continuous space representations and topic models with traditional vector-space models for ambiguity resolution. The selection of the optimal representation for specific lexical items and text types will be guided by the output of an ambiguity type detection mechanism, combined with genre and domain identification techniques. These parameters have up to now been left unexploited in favor of models that adopt a uniform approach (either topic-based or fine-grained) for handling different words and types of text. This is largely due to the difficulty of identifying the disambiguation needs of specific lexical items and texts, a challenge that MultiSem intends to address. The models that will be developed will be mainly data-driven and enriched with knowledge from large-scale semantic resources which have been shown to improve the performance of machine learning semantic processing methods. The combination of high-level ambiguity resolution techniques (topic models and neural networks) with fine-grained (vector-based) models, and the exploitation of the knowledge available in these resources will enhance the descriptive and processing capacities of the models. The research that will be conducted in MultiSem will renew the scientific perspectives in multilingual NLP, but also in linguistics and semantics due to the knowledge that will be extracted from large volumes of data. The proposed multi-layer ambiguity resolution models will also be exploited for improving lexical selection in translation applications. Lexical errors are found to be the predominant type of errors in automatically produced translations and could be avoided if Machine Translation (MT) systems were able to identify the meaning of words and larger textual units. By improving the quality of the generated translations, MultiSem will enhance the experience of numerous users of MT systems and will have an important social impact given the current pressing demand for quality processing of large volumes of digital content.

The ETAE project suggests significant advancements about the emergence and control of hydrodynamic instabilities in closed recirculating flows with a free surface. This generic flow configuration is present in numerous industrial contexts. The present aim is, from well-designed excitations by electro-active actuators, to manipulate the flow, and thus to identify the mechanisms promoting large-scale vortical instabilities arising in the presence of external mechanical noise. Bringing together the experimental/numerical skills on rotating flows at LIMSI, and the experience of GEEPS about modelling and conception of active actuators, will address important issues about the effect of parasitic noise on closed fluid systems. The exploratory side about actuators opens wide perspectives on the application of new measurement and control techniques in a fluid set-up, in close interaction with the development of new active materials expected to contribute in the future to fluid control strategies. The study of instabilities in closed rotating flows, triggered by rotating disks, has been one of the key topics for which LIMSI is internationally recognised. Such flows have now become classical topics due to their genericity and their importance in geophysical or industrial contexts. Using an experimental device consisting of a rotating vessel partially filled with liquid and a free surface, the team at LIMSI has shown evidence for instability modes due to the free surface. The flow before the instability is axisymmetric, and this axisymmetry is broken by instability modes above a given threshold (for the angular velocity of the disk). Two cases can be identified: weak deformations of the free surface where the instability manifests itself as a array of large-scale vortices, versus strong deformations where the free surface itself has broken its axisymmetry. From an experimental point of view, measurements of the free surface height in real time demands novel techniques. Besides, the strong deformation case remains even today a true challenge for numerical simulation. However, even in the weal deformation case, threshold measurements have revealed significant departures between experimental results and numerical predictions. Sensibility methods, developed only recently in the context of open flows, appear as relevant tools to understand the effect of generic external unsteadiness (of weak amplitude and mechanical origin) on the fluid system. Moreover, adding a perfectly controlled vibration to a given flow should explain, and more importantly reduce the mismatch between observed thresholds. Devising such actuators, the associated measurement methods, and integrating them into an efficient feedback loop represent as for today important technological challenges. This project is at the junction between active control of rotating flows at LIMSI and modelling of active-material-based actuators at GEEPS. Bringing together such skills is expected to lead to both fundamental and practical progress about the sensibility of confined flows to unavoidable parasitic vibrations. The exploratory side about actuators in the large deformation regime opens new important perspectives on the development of fluid-structure simulation codes as well as on the characterisation of electro-active materials.

More and more information on individuals (e.g., persons, events, biological objects) are available electronically in a structured or semi-structured form. However, selecting individuals satisfying certain constraints based on such data manually is a complex, error-prone, and time and personnel consuming effort. For this reason, tools that can automatically or semiautomatically answer questions based on the available data need to be developed. While simple questions can directly be expressed and answered using keywords in natural language, complex questions that can refer to type and relational information increase the precision of the retrieved results, and thus reduce the effort for posterior manual verification of the results. One example for this situation is the setting where electronic patient records are used to find patients satisfying non-trivial combinations of certain properties, such as eligibility criteria for clinical trials. Another example that will also be considered as a use case in this project is the setting where a student asks the examination office questions about study and examination regulations. In both cases, the original question is formulated in natural language. In the GoAsq project, we will investigate, compare, and finally combine two different approaches for answering questions formulated in natural language over textual, semi-structured, and structured data. One approach is the text-based question answering that directly answers natural language questions using natural language processing and information extraction techniques. The other tries to translate the natural language questions into formal, database-like queries and then answer these formal queries w.r.t. a domain-dependent ontology using database techniques. The automatic translation is required since it would be quite hard for the people asking the questions (e.g. medical doctors, students) to formulate them as formal queries. The ontology allows to overcome the possible semantic mismatch between the person producing the source data (e.g., the GPs writing the clinical notes) and the person formulating the question (e.g., the researcher formulating the trial criteria). GoAsq can thus leverage recent advances obtained in the ontology community on accessing data through ontologies, called ontology-based query answering (OBQA). More precisely, in Task 1 of the project we investigate the two use cases mentioned above (eligibility criteria; study regulations). In Task 2 we will introduce and analyze extensions to existing formal query languages that are required by these use cases. Task 3 will develop techniques for extracting formal queries from textual queries, and Task 4 will evaluate the approach obtained this way, compare it with approaches for text-based question answering, and develop a hybrid approach that combines the advantages of both.

Social media and other forms of online communication have triggered the emergence of new forms of written texts and increased the volume of multilingual user-generated content (UGC). Making these unlimited streams of non-canonical texts automatically understandable and actionable opens new scientific and social challenges. This is the main focus of the ParSiTi project. One of the most striking influences of social media on society is how they evolved to impact our perception of events. For instance, during the various Spring Revolutions, Facebook users were in the front line of the information war; more recently, during the November 2015 Paris Attacks, Twitter was used to gather information about the victims and to offer shelter to those stranded by these attacks. These events generated a steady flow of global textual interactions, crucially highlighting the lack of accurate tools to automatically process and understand these information streams. UGC, covering among others social networks, blogs and forums, differ from newspaper written languages, on which natural language processing (NLP) tools are most often trained and tested, in three important dimensions: (i) user-generated content is extremely diverse, rife with abbreviations, spelling mistakes, typographical and grammatical errors. It often lacks punctuation and mixes languages. In some cases, the spelling is akin to rough phonetisation. Added to a much richer variability, these phenomena hinder the performance of NLP pipelines. (ii) Overcoming English, the Web has now turned into a truly multilingual space. (iii) A strong contextualization as these non-canonical productions are tightly linked to contextual sources (videos, images, memes, game sessions, external URLs) and the inner nature of most social media encourages shorter sentences and threaded messages, which in turn favor the use of elliptical constructions. This leads to strong difficulties in rising ambiguities, for example in case of underspecified anaphoras, which complicate NLP tasks such as parsing or Machine Translation. The ParSiTi project aims at taking advantage of recent advances in statistical NLP and Deep Learning to address these challenges and improve access to multilingual user-generated content. We plan to design and release a fully integrated NLP software able to automatically process non-canonical texts in their context. To demonstrate the success of our approach, an accurate Machine Translation system, able to translate, in context, user-generated content between French, Arabic and English, will be developed. Such a system should prove valuable to researchers in linguistics, social sciences and for innovative private sector companies. Moreover, our software and data sets will be made freely available, so that they can be used for further work beyond the scope of the project, e.g. for information extraction or opinion mining. Developing this software is higly challenging and requires to push existing techniques to their limits, sometimes at the price of questioning assumptions that have long been taken for granted. The ParSiTi project will address three scientific challenges of increasing risk and complexity: (i) normalizing UGC and adapting parsing, along with translation models, to their peculiarities (ii) developing joint models to combine different sources of information without error propagation (iii) design context-aware models to cope with discussion anchored in a specific linguistic (e.g. comment in a threaded discussion) and extract-linguistic (e.g. images, URL, ...) contexts. ParSiTi will gather three partners: LIMSI for their expertise in Machine Translation and Deep Learning, LIPN for their expertise in joint models and parsing, and ALPAGE for their expertise in morpho-syntactic processing of social media, (deep) parsing, and out-of-domain adaptation.

Much literature has described the modern Western world (19th-21st Century) as predominately visual, marked by a “passion for seeing”. A wide array of studies have been devoted to the image, which sharply contrasts with the negligible research and reflection on the aural and listening — musical listening excepted. In our opinion, it is the study and practice of theatre that has suffered most spectacularly, and paradoxically, from this focus on the ocular: especially, if one considers how the Western stage has organized itself around an oral text, sometimes accompanied by (sung or instrumental) music, how the Greek stage rapidly took acoustics into account and how the Greek model has played a large part in theatre theory. Yet studies on theatre as an auditory place are largely absent – in contrast to the plethora of works on the performance’s visual aspects (scenography, stagecraft, lighting) and its relation to visual arts, painting and film. Hence, theatre studies seem based on the idea that the space of theatrical performance is organized principally by and for vision. Research effectuated on the orality of plays and on the actor’s voice has not impacted this model as they appear to be exiled in totally different theoretical fields. Thus, at the heart and root of the oversight of theatre’s sonic nature, one can foresee the overshadowing of the human "phônê", which is not only vocal, but verbal, as illustrated by Paul Zumthor’s description of “orality” in one of his last works (1994): a performed, possibly mediatized orality, engaging a socially significant aurality, because, as ethno-poetics well show, it finds its meaning in the event, but also because this event is in its very structure a linguistic event. Philosophers in particular have addressed the space given to the “invisible” in the world of images, notably through the status of listening, as well as of verbal expression and memory. The theatre, as discussed above, constitutes a privileged place and object for such an examination. The ECHO project emerged from this statement, as its interdisciplinary research team (from theatrical studies, acoustics, history of science, intermediality, ethnology, musicology, philosophy) adopts theoretical perspectives that break with dominant discourses focused either on (mute) images or on (vocal but non-verbal) bodies. From this new perspective, Western theatre is an acoustic space organized by and for the “spoken” voice. This definition was embraced by modern theatre acousticians of the turn of 20th Century, then questioned and rearticulated according to transformations in the representation of phonation (its production and its acoustics) due to evolutions of medical, linguistic, psychoanalytic knowledge, and acoustic science and to developments of new sound technologies—microphones, phonographs, telephones, loudspeakers. Such a perspective unlocks and opens a theoretical space that allows for a dual study of stage practice and the spoken voice. Thus, one aspect of our project consists in examining “modern scene” creations (1950-2000), documented by Giovanni Lista’s canonic survey, and describing the genesis of novel oral and aural forms related to new phonic worlds and modes of vision. This will be achieved mainly through reconstituting the acoustic history of two auditoriums, developing listening analyses of theatre audio archives (just as the “theatre of images” is full of textual echoes), and conducting surveys with audience members of past performances. ECHO will, thus, contribute to the rewriting of Western theatre’s recent history as well as to the elaboration, through the case of theatrical performance, of a truly modern theorization of orality – one that is freed from the two current seductive and reductive dominating models: the vocalized ritual and the technologized stage.