CALL FOR PARTICIPATION

IBERLEF 2023 Task - FinancES. Financial Targeted Sentiment Analysis in Spanish

Held as part of iberLEF 2023 https://sites.google.com/view/iberlef-2023, a shared evaluation campaign for Natural Language Processing (NLP) systems in Spanish and other Iberian languages

September 26th 2023, Jaen

Codalab link: https://codalab.lisn.upsaclay.fr/competitions/10052

Dear All,

We are inviting researchers and students to participate in the shared-task FinancES. Financial Targeted Sentiment Analysis in Spanish held as part of iberLEF 2023, shared evaluation campaign for Natural Language Processing (NLP) systems in Spanish and other Iberian languages.

This shared task aims to explore targeted sentiment analysis in the financial domain. Specifically, the approach adopted here is grounded in the field of microeconomics. In this regard, Bowles (2004) explains the role of economic agents, that is to say, individuals or organizations impacting the economy. The author states that the main microeconomic agents in the capital market are consumers (households/individuals), companies (firms), governments, and central banks. Consequently, in order to develop a sentiment analysis method where different viewpoints are considered, three different perspectives are included: (1) economic target of the news item; (2) individual economic agent: companies; and (3) individual economic agent: consumers –the target is the sector where the economic fact applies, and companies produce the goods and services that households/individuals consume. From these three viewpoints, the news item has an impact on the target and the economic agents which are considered as positive, negative, or neutral. With all, two tasks are proposed. On the one hand, a task combining the challenges of aspect-term extraction for identifying the target entity in text, and aspect-based sentiment classification for determining the sentiment polarity towards the target. On the other hand, a task devoted to assessing the impact of a news headline on both other economic agents, namely, companies and consumers.

The participants will be provided development, development_test, training and test datasets in Spanish. The dataset for this task is composed of news headlines written in Spanish collected from digital newspapers specialized in economic, financial and political news. The dataset is labeled with the target entity and the sentiment polarity on three dimensions: target, companies, and consumers. That is, given a headline, it has been manually classified as positive, neutral, or negative for three specific entities: (1) target entity (i.e., the specific company or asset where the economic fact applies), (2) companies (i.e., the entities producing the goods and services that others consume), and (3) consumers (i.e., households/individuals). Each headline was annotated by three members of the organization committee. In case of disagreement, the annotators discussed the special case and, if no agreement was reached, the headline was discarded. During this first step, we compiled about 14k headlines, the headlines with a short length or those that did not specify a target entity were filtered out. The final dataset is composed of 8k-10k news headlines. For the shared tasks, training and test sets will be released (80%-20%).

In order to facilitate participation in the competition, a Google Colab notebook has also been provided. In this notebook, it is shown how to load the development dataset and how to train a baseline for both tasks based on SpaCy for the identification of the main economic target and a Bag-of-Words (BoW) with linear regression of the polarity of each dimension. In addition, it is shown how to calculate the final F1-score of each task and how to generate the final submission file. To download the data, the notebook and participate, go to https://codalab.lisn.upsaclay.fr/competitions/10052#participate.

Best regards,

The FinancES 2023 organizing committee

References

-

Bowles, S. (2004). Microeconomics: Behavior, institutions, and evolution. Princeton University Press.

Important dates

-

Release of development corpora: Feb 13, 2023 -

Release of training corpora: Mar 13, 2023 -

Release of test corpora and start of evaluation campaign: Apr 17, 2023 -

End of evaluation campaign (deadline for runs submission): May 3, 2023 -

Publication of official results: May 5, 2023 -

Paper submission: May 28, 2023 -

Review notification: Jun 16, 2023 -

Camera ready submission: Jul 6, 2023 -

IberLEF Workshop (SEPLN 2023): Sep 26, 2023 -

Publication of proceedings: Sep ??, 2023

Organizing committee

-

José Antonio García-Díaz (UMUTeam,, Universidad de Murcia) -

Ángela Almela Sánchez-Lafuente (UMUTeam, Universidad de Murcia) -

Francisco García-Sánchez (UMUTeam, Universidad de Murcia) -

Gema Alcaraz-Mármol (UMUTeam, Universidad de Castilla La Mancha) -

María José Marín (UMUTeam, Universidad de Murcia) -

Rafael Valencia-García (UMUTeam, Universidad de Murcia)