Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

EFSA: Towards Event-Level Financial Sentiment Analysis

Tianyu Chen, Yiming Zhang, Guoxin Yu, Dapeng Zhang, Li Zeng, Qing He, Xiang Ao

TL;DR

This paper reconceptualizes the event extraction as a classification task by designing a categorization comprising coarse-grained and fine-grained event categories and formulation of the four-hop Chain-of-Thought LLM-based approach for this task.

Abstract

In this paper, we extend financial sentiment analysis~(FSA) to event-level since events usually serve as the subject of the sentiment in financial text. Though extracting events from the financial text may be conducive to accurate sentiment predictions, it has specialized challenges due to the lengthy and discontinuity of events in a financial text. To this end, we reconceptualize the event extraction as a classification task by designing a categorization comprising coarse-grained and fine-grained event categories. Under this setting, we formulate the \textbf{E}vent-Level \textbf{F}inancial \textbf{S}entiment \textbf{A}nalysis~(\textbf{EFSA} for short) task that outputs quintuples consisting of (company, industry, coarse-grained event, fine-grained event, sentiment) from financial text. A large-scale Chinese dataset containing $12,160$ news articles and $13,725$ quintuples is publicized as a brand new testbed for our task. A four-hop Chain-of-Thought LLM-based approach is devised for this task. Systematically investigations are conducted on our dataset, and the empirical results demonstrate the benchmarking scores of existing methods and our proposed method can reach the current state-of-the-art. Our dataset and framework implementation are available at https://anonymous.4open.science/r/EFSA-645E

EFSA: Towards Event-Level Financial Sentiment Analysis

TL;DR

This paper reconceptualizes the event extraction as a classification task by designing a categorization comprising coarse-grained and fine-grained event categories and formulation of the four-hop Chain-of-Thought LLM-based approach for this task.

Abstract

In this paper, we extend financial sentiment analysis~(FSA) to event-level since events usually serve as the subject of the sentiment in financial text. Though extracting events from the financial text may be conducive to accurate sentiment predictions, it has specialized challenges due to the lengthy and discontinuity of events in a financial text. To this end, we reconceptualize the event extraction as a classification task by designing a categorization comprising coarse-grained and fine-grained event categories. Under this setting, we formulate the \textbf{E}vent-Level \textbf{F}inancial \textbf{S}entiment \textbf{A}nalysis~(\textbf{EFSA} for short) task that outputs quintuples consisting of (company, industry, coarse-grained event, fine-grained event, sentiment) from financial text. A large-scale Chinese dataset containing news articles and quintuples is publicized as a brand new testbed for our task. A four-hop Chain-of-Thought LLM-based approach is devised for this task. Systematically investigations are conducted on our dataset, and the empirical results demonstrate the benchmarking scores of existing methods and our proposed method can reach the current state-of-the-art. Our dataset and framework implementation are available at https://anonymous.4open.science/r/EFSA-645E
Paper Structure (23 sections, 5 figures, 4 tables)

This paper contains 23 sections, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Problem Formulation and Discussion
  3. Problem Formulation
  4. The Sub-tasks of EFSA
  5. Relatedness to ABSA Task
  6. Dataset
  7. Data Collection
  8. Data Annotation
  9. Data Distribution
  10. Experimental Evaluation
  11. Benchmark Methods
  12. Our Framework
  13. Evaluation Metrics
  14. Experimental Results
  15. Main Results and Analysis
  16. ...and 8 more sections

Figures (5)

  • Figure 1: An example of Event-level Financial Sentiment Analysis from financial news. There could be multiple entities associated with their events with different sentiments.
  • Figure 2: The statistics of data distribution of our dataset. (a) presents a nested pie chart that illustrates the distribution of event labels. The inner layer and the outer layer respectively represent the distribution of coarse- and fine-grained events; the dark and light shades of the same color represent the coarse-grained event and its subdivided fine-grained events. (b), (c), (d), and (e) present the distribution among different publish times, overall sentiments, news body length, and sentiment distribution of each coarse-grained event. (f) presents the distribution of industries, where various colors denote $32$ distinct industries. The size of the color blocks in (a) and (f) represents the data size.
  • Figure 3: An illustration of the four-hop CoT framework. E1, E2, S respectively denotes the set of coarse-grained events, fine-grained events and sentiment polarities, respectively.
  • Figure 4: Examples provided include the input news text, the corresponding gold labels, and the predicted quantities. The red font denotes the incorrect part of the prediction.
  • Figure 5: Screenshot of the annotation system.