4 Available PhD Positions:

  1. Subject: How Specific Sentiment Features Can Impact Social Network Analysis?
  2. Subject: Don't do it: Social Web Mining for suicide prevention
  3. Subject: Active Learning classification for spatially and auto-correlated data: Application to remote sensing data
  4. Subject: Data Mining and Information Visualization for Facilitating the Medical Analysis of Patient Movement
  1. PhD position available in Montpellier, France

    PDF Subject: How Specific Sentiment Features Can Impact Social Network Analysis?

    Description: Sentiment analysis is a "hot" topic for Natural Language Processing (NLP) and data mining. Major conferences in these fields propose specific sessions about this crucial topic to evaluate "people's pulse". But currently, the methods are often based on basic analysis with statistics concerning the "positive" and "negative" words in documents. More sophisticated methods are proposed, but, in general, they are not integrated in a global system which takes into account several areas (e.g. NLP, data mining, communication). Moreover, these methods are often adapted to one language and/or specific kinds of texts (movies reviews, product reviews, blogs, tweets, and so forth).

    In our project, we propose a generic method to develop a method for fine-grained analysis of texts. A more subtle analysis will be possible by the use of new media communication skills and the acquisition of a more precise sentiment model. The precision of the model that we propose is given by (1) a complete emotion classification, (2) a sentiment analysis relative to a topic or a sub-topic from a text. This work will allow us to analyze modern communication media (tweets, blogs, etc.) which use a very specific and evolving vocabulary. Our automatic methods will be based on seed knowledge, enriched incrementally. So, the development of our approaches will be independent of the domain and of the language (French, Spanish, English, etc.).

    We will focus our work on Environment domain. Actually, this topic represents a priority for the research in Montpellier (Pôle de Compétitivité) and for the NUMEV Labex (leader: LIRMM) and GEOSUD Equipex (leader: TETIS). For instance NUMEV seeks to harmonize the approaches of "hard" sciences and life and environmental sciences in order to pave the way for an emerging interdisciplinary group with an international profile.

    Institutions: University of Montpellier 2, LIRMM (Laboratory of Informatics, Robotics, and Microelectronics in Montpellie), TETIS.

    Supervisors: Pascal Poncelet and Mathieu Roche.

    How to Apply: For additional information about the position, please contact PhD. candidate Juan Antonio LOSSIO-VENTURA, juan.lossio@lirmm.fr. An ideal candidate must be from Peru (mandatory), should have a Master's degree (not mandatory) in computer science or mathematics. Also those who are expected to finish their Master's degree may apply. Good programming skills (e.g. Matlab, Java, C++) and fluency in spoken and written English and/or French are required. The starting date is negotiable. Applications (including a CV, academic records, motivation letter and appreciation letters if available) are to be sent to the following people:

    Application Deadline: The application deadline is November 5, 2014.

    Read more: Complete subject.

  2. PhD position available in Montpellier, France

    PDF Subject: Don't do it: Social Web Mining for suicide prevention

    Description: Suicide is the deliberate act of ending his own life. Suicide reveals serious personal problems but also often reflect a deterioration of the social context in which an individual lives. Risk factors are multiple and complex (changes in personal relationships, harassment, addiction, unemployment, clinical depression and other forms of mental illness, etc.). According to a very recent and alarming WHO report (September 4, 2014), one person commits suicide every 40 seconds - more than all the yearly victims of wars and natural disaster. The number of suicides in the world in 2012 is estimated at 804,000, which represents an age-standardized rate of suicide to 11.4 per 100 000 (15 males and 8 females). Europe is far from being spared with an age-standardized suicide rate of 12 per 100,000 inhabitants in 2012 (20 years for men and 4.9 for women). The highest suicide rates are measured in the Baltics and Central Europe (Estonia, Hungary, Latvia, Lithuania and Slovenia, with more than 17 deaths per 100,000 inhabitants). The majority of people who commit suicide are men, by a ratio of two compare to women, as they tend to use less effective methods. Suicide is also related to age. People under the age of 25 and over 50 are particularly at risk. Most suicide attempts are supported by hospital ER. Suicide is a major public health with strong socio-economic consequences. As part of the Action Plan for the 2013-2020 Mental Health, WHO Member States have pledged to meet the global target of 10% reduction in suicide rates all over the world by 2020.

    The main objective of this thesis is to develop new technologies for early identification of at-risk individuals through their use of the social web. S. Bringay and J. Azé (supervisors) have already developed an initial prototype based on a predictive model, which has proven to be effective for identifying at-risk people through their publications on the Twitter social network. In this thesis, we extend the problem to other media (blogs, forums, social-Facebook, email, instant messaging, etc.) and focus on two scenarios: 1) the semi-automatic detection model of suicidal profiles will be used by psychiatrists, to monitor patients on the Facebook or Twitter network, who have stayed in their service after a first suicide attempt. We intend to capture a possible deterioration in their mental state in order to offer them assistance when needed. The aim is to prevent suicide recurrence; 2) the semi-automatic detection model of suicidal profiles will be used by organizations as Arrêt Demandé International receiving contacts via instant messaging and emails to help them prioritize their responses.

    The work of the PhD student will focus on three major challenges: 1) use methods of information retrieval to collect heterogeneous texts issued from the social web and dealing with topics associated with suicide (depression, anorexia, etc.). These data will be stored in a texts warehouse, robust to the characteristics of the data (heterogeneity, large volumes, velocity, etc.); 2) tag at-risk messages with a new typology involving various risk factors and use these marks to mine the messages and detect at-risk people. The objective is to obtain a predictive model to effectively trigger graduated answers; 3) instantiation of model for the two scenarios previously identified.

    Institutions: University of Montpellier 2, LIRMM (Laboratory of Informatics, Robotics, and Microelectronics in Montpellie).

    Supervisors: Jérôme Azé and Sandra Bringay.

    How to Apply: For additional information about the position, please contact PhD. candidate Juan Antonio LOSSIO-VENTURA, juan.lossio@lirmm.fr. An ideal candidate must be from Peru (mandatory), should have a Master's degree (not mandatory) in computer science or mathematics. Also those who are expected to finish their Master's degree may apply. Good programming skills (e.g. Matlab, Java, C++) and fluency in spoken and written English and/or French are required. The starting date is negotiable. Applications (including a CV, academic records, motivation letter and appreciation letters if available) are to be sent to the following people:

    Application Deadline: The application deadline is November 5, 2014.

    Read more: Complete subject.

  3. PhD position available in Montpellier, France

    PDF Subject: Active Learning classification for spatially and auto-correlated data: Application to remote sensing data

    Description: The study of the spatial dynamics of regional or national scales leads to cross multi-source data of various types maps, surveys, sensor data or satellite images. These data carry information of spatial, temporal and thematic character. The study of satellite images allows automated explorative analysis of the territories. Once the images are acquired, a process of segmentation is performed to extract the objects of interest. The segmentation step is a critical phase that requires a certain expertise, both on the tool used and the nature of the land concerned. This phase is a key to the success of the analysis of the land remote sensing process. Once the segmentation carried out, a set of objects is obtained, whose number depends on the granularity associated with the study to be conducted. The ultimate goal is to obtain a classification of these objects to produce as close as possible to the reality of land use mapping.

    The number of objects produced during segmentation can be huge (between 5000 and 50000 objects), for this reason it is not always possible to ask the experts to validate all these items manually. This problem of Object Oriented Classification (OOC) involves the automatic classification of these segment obtained by the satellite image in order to produce a detailed map of the study area. In order to automatically perform this operation we need to train a predictive model with a labeled training set. Most of the time, these methods assume that a big set of example, already labeled, exist and it is easily ready to be employed. The construction of this training set has a cost, as the expert must analyze each object one by one and assign a class of land. Once this step is completed, the classification algorithm can then be driven.

    Starting from all the previous aspects, during this thesis we would to focus the PhD efforts to improve current techniques for the OOC of satellite images considering active learning techniques to reduce the cost of the labeling process and involving spatial information that really characterized our domain application.

    Institutions: University of Montpellier 2, LIRMM (Laboratory of Informatics, Robotics, and Microelectronics in Montpellie).

    Supervisors: Dino Ienco and Maguelonne Teisseire.

    How to Apply: For additional information about the position, please contact PhD. candidate Juan Antonio LOSSIO-VENTURA, juan.lossio@lirmm.fr. An ideal candidate must be from Peru (mandatory), should have a Master's degree (not mandatory) in computer science or mathematics. Also those who are expected to finish their Master's degree may apply. Good programming skills (e.g. Matlab, Java, C++) and fluency in spoken and written English and/or French are required. The starting date is negotiable. Applications (including a CV, academic records, motivation letter and appreciation letters if available) are to be sent to the following people:

    Application Deadline: The application deadline is November 5, 2014.

    Read more: Complete subject.

  4. PhD position available in Montpellier, France

    PDF Subject: Data Mining and Information Visualization for Facilitating the Medical Analysis of Patient Movement

    Description: Visual analytics is the science of analytical reasoning facilitated by interactive visual interfaces. More precisely, "visual analytics combines automated analysis techniques with interactive visualizations for effective understanding, reasoning and decision making on the basis of very large and complex datasets". Automated analysis techniques include statistics, mathematics, knowledge representation, management, and discovery technologies. In this project, we focus on combining data mining techniques with information visualization.

    Recent technologies such as Kinect, Wiiboard, Tablet, Joypad, eye tracking devices, etc., enable to record patient's behavior. For medicine, analyzing this behavior is a key to establish efficient diagnostics and better design recovery good practices. To perform a strong diagnostic, data generated by recording devices (body movements, eye movements, etc.) has to be analyzed taking account of the patient background (health problem, personal information, etc.). In this context, therapists are confronted to large amount of heterogeneous data and their analyses require automatic methods. The difficulty lies not in generating but rather in sifting through this mass raw of data in order to discover patterns. This thesis focuses on designing data mining and information visualization techniques to create interfaces helping healthcare professionals to understand the data.

    A number of issues associated with the design of such interfaces will be examined in the context of this thesis in order to provide therapists with the tools necessary to perform their analyses (see for example the recent promising work of Bernard et al). It is of utmost importance that these interfaces be as simple and intuitive as possible, as a misinterpretation of the data due to imperfect interface design could hamper the rehabilitation process. It is for this reason that an in-depth study of knowledge discovery methods must be undertaken by experts in data mining, information visualization and human-computer interaction, in close collaboration with therapists. In order to design the interface, we will follow the process described by Tamara Munzner (domain problem characterization, data/operation abstraction design, encoding/interaction technique design, algorithm design). Each of these steps will be validated using the article's recommendations. Next user testing will be performed to verify the effectiveness of our approach, using the experimental protocols described by Helen Purchase.

    Institutions: University of Montpellier 2, LIRMM (Laboratory of Informatics, Robotics, and Microelectronics in Montpellie).

    Supervisors: Jérôme Azé and Arnaud Sallaberry .

    How to Apply: For additional information about the position, please contact PhD. candidate Juan Antonio LOSSIO-VENTURA, juan.lossio@lirmm.fr. An ideal candidate must be from Peru (mandatory), should have a Master's degree (not mandatory) in computer science or mathematics. Also those who are expected to finish their Master's degree may apply. Good programming skills and fluency in spoken and written English and/or French are required. The starting date is negotiable. Applications (including a CV, academic records, motivation letter and appreciation letters if available) are to be sent to the following people:

    Application Deadline: The application deadline is November 5, 2014.

    Read more: Complete subject.

Links

Laboratory of Informatics, Robotics, and Microelectronics of Montpellier







Complements

 
First design provided by Free Web Templates - your source for free website templates