(May 2022 - April 2024)
Social media language analysis for early detection of mental health disorders
(FAPESP 2021/08213-0)
The observation that individuals with mental health disorders such as depression and anxiety are often regular users of social media has led to the development of a wide range of studies in the Natural Language Processing (NLP) field for risk assessment based on the kind of language employed by these individuals. Existing work in the field is however largely dedicated to the English language, and tends to consider publications (e.g., tweets) produced at any time, including even those produced after the individual is already clinically diagnosed. Thus, models of this kind tend to focus more on the issue of distinguishing individuals with and without a certain disorder, but are perhaps less able to anticipate it as a means to prevent its possible aggravation. Based on these observations, this project proposes to explore the temporal information provided by the Twitter platform for the study and development of computational models for early recognition of depression and anxiety disorder in Portuguese using a database - called the SeptemberBR corpus - designed so as to include only texts that are chronologically prior to the date of diagnosis reported by social media users. A study of this kind, in addition to introducing a novel (and possibly more useful) formulation of the present computational problem, opens up the opportunity for a number of scientific contributions in the NLP field, including the modeling of textual and non-textual features and the use of recent neural learning methods, and enables novel solutions for a pressing issue of great social interest.
Details are available from the project page.
(April 2019 - April 2020)
Moral stance recognition from text
(USP PRP 668/2018)
Stance recognition from text is a well-known NLP research topic. Systems of this kind aim to determine whether the author of a text may have a favourable, contrary or neutral attitude (or stance) towards a particular target (e.g., a person, a company, a product, a piece of legislation etc.) Computational models of stance recognition are common in the case of the English language, but much less so in the case of Portuguese. Based on these observations, this project intends to develop a prototype of a system for the recognition of moral and/or ideological stances from Portuguese texts by making use of supervised machine learning methods, whose initial results will be taken as the basis for future studies in this field.
Details are available from the project page.
(March 2017 - February 2019)
Computational models of human personality for Natural Language Processing Applications
(FAPESP grant 2016/14223-0)
The project focuses on the computational treatment of human personality from both natural language understanding and generation perspectives. Based on the well-established Big Five model of personality traits, the project proposes the collection and annotation of a basic linguistic-computational resource - which can be seen as a parallel corpus of texts and personality inventories - in order to map the relationships between personality traits and a broad range of linguistic phenomena, and then use this resource to build both computational models of personality recognition from text, and models of text generation based on target personality profiles.
Details are available from the project page.
(June 2014 - May 2016)
Generating Natural Language Descriptions in Visual Contexts
(FAPESP grant 2013/23169-0)
The project concerns the interaction between language production and images. Our aim is to collect and annotate corpora of descriptions in visual domains, and then use these resources to build novel computational models of reference production.
Details are available from the project page.
(Aug 2011 - Feb 2012)
Referring Expression Generation in Interactive 3D Domains
(CAPES grant 6415/10-5).
The project concerns the extension of previous work on the use of logically redundant information in referring expressions generation as a means to facilitate search in large and/or complex domain structures such as spatial domains.
(Sept 2009 – Aug 2011)
Hybrid Models of Surface Realisation
(FAPESP grant 2009/08499-9).
The project applies the traditional 2-stage generation approach to the problem of text generation for morphologically-rich languages such as Portuguese, combining standard statistical language models with corpus knowledge.
(Nov 2007 – Oct 2009)
Resolution and Interpretation of Referring Expressions in Text Generation
(CNPq grant 484015/2007 9)
(Nov 2006 – Oct 2008)
Building a Parallel Corpus with Annotated Co-reference Chains for Statistical Machine Translation
(FAPESP grant 2006/03941-7)
(Sept 1998 - July 2003)
Generating references in hierarchical domains: the case of document deixis.
PhD project at the University of Brighton
(CNPq grant 200123/98-0)