Participation at TASS 2012 is closed.
Participation
Participants are required to register for the task(s) in order to obtain the corpus.
Registration must be done by mail to tass AT daedalus.es.
Submissions
Experiments
Experiments must be submitted in a plain text file with the following format:
twitit \t polarity \t topic
where twitid is the twit ID for every message in the test corpus, the polarity contains one of the 6 valid tags (P+, P, NEU, N, N+ and NONE), and the same for topic.
Although the polarity level must be classified into those levels and the results will be evaluated for the 5 of them, the evaluation results will include metrics that consider just 3 levels (POSITIVE, NEUTRAL and NEGATIVE).
Participants may submit results for one or both tasks.
Submissions for any of the tasks must be done by email to tass AT daedalus.es.
Reports
We invite participants to submit a report describing the experiments and discussing the results.
Papers should follow the usual SEPLN template given in the author guidelines page. Reports can be written in Spanish or English. In this case there is no limitation in extension as they will be included in the electronic working notes of the conference.
Submissions must be done by email to tass AT daedalus.es and the deadline is August 19th, 2012.
Results
15 groups registered and finally 9 groups sent their submissions. Results are listed in the tables below.
The detailed results for each run, the relevance judgement files (the correct labels for the test set) and a PHP script for performing your own evaluation can be found in the Downloads section of the Corpus description page.
Task 1. Sentiment Analysis (5 levels + NONE)
Run Id | Group | Precision |
pol-elhuyar-1-5l | Elhuyar Fundazioa | 65.29% |
pol-l2f-1-5l | L2F - INESC | 63.37% |
pol-l2f-3-5l | L2F - INESC | 63.27% |
pol-l2f-2-5l | L2F - INESC | 62.16% |
pol-atrilla-1-5l | La Salle - Universitat Ramon Llull | 57.01% |
pol-sinai-4-5l | SINAI - UJAEN | 54.68% |
pol-uned1-2-5l | LSI UNED (tamara & jorge) | 53.82% |
pol-uned1-1-5l | LSI UNED (tamara & jorge) | 52.54% |
pol-uned2-2-5l | LSI UNED 2 (angel, juan manuel & ana) | 40.41% |
pol-uned2-1-5l | LSI UNED 2 (angel, juan manuel & ana) | 39.98% |
pol-uned2-3-5l | LSI UNED 2 (angel, juan manuel & ana) | 39.47% |
pol-uned2-4-5l | LSI UNED 2 (angel, juan manuel & ana) | 38.59% |
pol-imdea-1-5l | IMDEA | 36.04% |
pol-sinai-2-5l | SINAI - UJAEN | 35.65% |
pol-sinai-1-5l | SINAI - UJAEN | 35.28% |
pol-sinai-3-5l | SINAI - UJAEN | 34.97% |
pol-tudelft-5y6-5l | TUDELFT | 34.05% |
pol-tudelft-3y4-5l | TUDELFT | 33.46% |
pol-tudelft-1y2-5l | TUDELFT | 33.04% |
pol-uma-1-5l | UMA | 16.73% |
Task 1. Sentiment Analysis (3 levels + NONE)
Run Id | Group | Precision |
pol-elhuyar-1-3l | Elhuyar Fundazioa | 71.12% |
pol-l2f-1-3l | L2F - INESC | 69.05% |
pol-l2f-3-3l | L2F - INESC | 69.04% |
pol-l2f-2-3l | L2F - INESC | 67.63% |
pol-atrilla-1-3l | La Salle - Universitat Ramon Llull | 61.95% |
pol-sinai-4-3l | SINAI - UJAEN | 60.63% |
pol-uned1-1-3l | LSI UNED (tamara & jorge) | 59.03% |
pol-uned1-2-3l | LSI UNED (tamara & jorge) | 58.77% |
pol-uned2-1-3l | LSI UNED 2 (angel, juan manuel & ana) | 50.08% |
pol-imdea-1-3l | IMDEA | 45.95% |
pol-uned2-2-3l | LSI UNED 2 (angel, juan manuel & ana) | 43.61% |
pol-tudelft-5y6-3l | TUDELFT | 43.61% |
pol-uned2-4-3l | LSI UNED 2 (angel, juan manuel & ana) | 41.20% |
pol-uned2-3-3l | LSI UNED 2 (angel, juan manuel & ana) | 40.43% |
pol-tudelft-3y4-3l | TUDELFT | 40.27% |
pol-tudelft-1y2-3l | TUDELFT | 38.45% |
pol-uma-1-3l | UMA | 37.61% |
pol-sinai-2-3l | SINAI - UJAEN | 35.83% |
pol-sinai-1-3l | SINAI - UJAEN | 35.58% |
pol-sinai-3-3l | SINAI - UJAEN | 35.11% |
Task 2. Trending topic coverage
Run Id | Group | Precision |
top-l2f-2 | L2F - INESC | 65.37% |
top-l2f-1y3 | L2F - INESC | 64.92% |
top-atrilla-1 | La Salle - Universitat Ramon Llull | 60.16% |
top-tudelft-135 | TUDELFT | 56.27% |
pol-uned2-5a8 | LSI UNED 2 (angel, juan manuel & ana) | 45.26% |
top-imdea-1 | IMDEA | 45.24% |
top-tudelft-246 | TUDELFT | 43.72% |
pol-uned2-9a12 | LSI UNED 2 (angel, juan manuel & ana) | 42.24% |
pol-uned2-1a4 | LSI UNED 2 (angel, juan manuel & ana) | 40.51% |
top-sinai-5 | SINAI - UJAEN | 39.37% |
top-sinai-4 | SINAI - UJAEN | 37.79% |
top-sinai-2 | SINAI - UJAEN | 34.76% |
top-sinai-3 | SINAI - UJAEN | 34.06% |
top-sinai-1 | SINAI - UJAEN | 32.34% |
pol-uned1-1y2 | LSI UNED (tamara & jorge) | 30.98% |
Reports
- TASS: Detecting Sentiments in Spanish Tweets [322.8KB, 2013-04-29]
Xabier Saralegi Urizar, Iñaki San Vicente Roncal
Elhuyar Fundazioa - The L2F Strategy for Sentiment Analysis and Topic Classification [499.8KB, 2013-04-29]
Fernando Batista, Ricardo Ribeiro
L2F - INESC-ID, ISCTE-IUL - Lisbon University Institute - Sentiment Analysis of Twitter messages based on Multinomial Naive Bayes [118.7KB, 2013-04-29]
Alexandre Trilla, Francesc Alías
Grup de Recerca en Tecnologies Mèdia - La Salle, Universitat Ramon Llull - UNED at TASS 2012: Polarity Classification and Trending Topic System [245.5KB, 2013-04-29]
Tamara Martín-Wanton, Jorge Carrillo de Albornoz
UNED NLP & IR Group - UNED @ TASS: Using IR techniques for topic-based sentiment analysis through divergence models [157KB, 2013-04-29]
Angel Castellano González, Juan Cigarrán Recuero, Ana García Serrano
UNED - Lexicon-Based Sentiment Analysis of Twitter Messages in Spanish [405.9KB, 2013-04-29]
Antonio Moreno-Ortiz, Chantal Pérez-Hernández
Universidad de Málaga