The main goal of OPTIMISM ("Opinion Text Mining from Social Meshes") is the development of an opinion mining system for the detection of opinions, in quasi real-time, about relevant political actors, including politicians and political parties. Opinion-rich texts will be gathered from the Portuguese social media, and classified according to their semantic orientation and intensity.
Note: we have received funding for the new Reaction project and we are contributing the work on mining resources initiated under OPTIMISM into that new development.
With the increasing availability of user-generated contents, such as blogs, Internet forums and social networks, users have more opportunities to express their opinions and make them available to everyone. Publicly available opinions provide valuable information for decision-making processes based on a new collective intelligence paradigm designated as “crowdsourcing.”
OPTIMISM selectively harvests and then mines multifaceted texts available on news media and the social web. The texts include political weblog posts and comments to news published in the web sites of mainstream media, an extremely rich and dynamic source of subjective information. The opinion mining system prototype under development will use text pattern classifiers trained with examples (annotated corpora) and rules (lexica and grammars) from large-scale and fine-grained linguistic resources, specifically conceived for Portuguese.
The OPTIMISM project focuses on a challenging interdisciplinary problem that demands the application of knowledge and methods from information retrieval, text mining, computational linguistics, and political and social sciences. The main challenge is the detection of trend changes in the electoral behavior ahead of existing polling technology by mining the social media websites, investigating how these trends correlate with those obtained using conventional polling methods.
OPTIMISM will use text pattern classifiers trained with examples (annotated corpora) and rules (lexica and grammars) from large-scale and fine-grained linguistic resources, specifically conceived for Portuguese.
Current activity is centred on the development of resources, including:
- a sentiment lexicon for Portuguese, made up of simple and multiword lexical units
- syntactic-semantic rules describing the cases of polarity assigned to predicates at the lexical level that can be reversed depending on the syntactic construction
- an annotated corpus that corresponds to a representative collection of opinion-labeled texts from the social media
- SFRH/BPD/45416/2008, Post-doctoral research scholarship for Paula Carvalho, from 1-Dez-08 to 31-Oct-2010
- SFRH/BD/65972/2009, Doctoral research scholarship for Carlos Costa, from 1-Dez-10 to 31-12-2010
An interdisciplinary group with common and complementary interests and past relevant research, from XLDB@LASIGE, LIACC and ICS, congregates the required competencies to face that challenge.
From ICS and LIACC/FEUP:
Mário J. Silva, Paula Carvalho, Carlos Costa, Luís Sarmento, Automatic Expansion of a Social Judgment Lexicon for Sentiment Analysis Technical Report. TR 10-08. University of Lisbon, Faculty of Sciences, LASIGE, December 2010. doi: 10455/6694
Mário J. Silva, Paula Carvalho, Luís Sarmento, Eugénio Oliveira, Pedro Magalhães, The Design of OPTIMISM, an Opinion Mining System for Portuguese Politics.New Trends in Artificial Intelligence: Proceedings of EPIA 2009 - Fourteenth Portuguese Conference on Artificial Intelligence p. 565-576, October, 2009. Universidade de Aveiro.
Paula Carvalho, Luís Sarmento, Mário J. Silva, Eugénio de Oliveira, Clues for Detecting Irony in User-Generated Contents: Oh...!! It's “so easy“ ;-).Proceedings of TSA'09 - Text Sentiment Analysis, held at ACM CIKM 2009 Hong Kong, China, November, 2009. ACM Press.
Luís Sarmento, Paula Carvalho, Mário J. Silva, Eugénio de Oliveira, Automatic Creation of a Reference Corpus for Political Opinion Mining in User-Generated Content.Proceedings of TSA'09 - 1st International CIKM Workshop on Topic-Sentiment Analysis for Mass Opinion Measurement, held at ACM CIKM 2009 Hong Kong, China, November, 2009. ACM Press.