Financial Opinion Mining

In this tutorial, we will show where we are and where we will be to those researchers interested in this topic. We divide this tutorial into three parts, including coarse-grained financial opinion mining, fine-grained financial opinion mining, and possible research directions. This tutorial starts by introducing the components in a financial opinion proposed in our research agenda and summarizes their related studies. We also highlight the task of mining customers’ opinions toward financial services in the FinTech industry, and compare them with usual opinions. Several potential research questions will be addressed. We hope the audiences of this tutorial will gain an overview of financial opinion mining and figure out their research directions.


Goal of the Tutorial
When it comes to financial opinion mining, bullish and bearish come into people's minds. However, more fine-grained information will be missed if we only focus on the market sentiment analysis of financial documents. Thanks to the recent "CS + X" trend, more interdisciplinary cooperation exists between computer science and other domains. In the "NLP + Finance" community, lots of recent works pay their attention to in-depth analysis of different kinds of financial documents rather than market sentiment prediction. For example, our previous works (Chen et al., 2018(Chen et al., , 2019a find that the numeral information extracted from financial social media data is comparable to the price targets extracted from professional analysts' reports. Keith and Stent (2019) analyze the pragmatic and semantic features in the earnings conference calls and discuss how these features influence the investor's decision-making process. Zong et al. (2020) point out the difference between the textual factors and cognitive factors by comparing the accurate and inaccurate professional analysts' reports. The abovementioned works conclude the necessity of capturing fine-grained opinions in the financial narratives. As the increasing interest of our community on this topic, recently, more and more related workshops spring up in the leading conferences, including FinWeb-2021 in the Web Conference, FinNLP-2021in IJCAI, FinIR-2020in SIGIR, and FNP-2020 In this tutorial, we will show where we are and where we will be to those researchers interested in this topic. We divide this tutorial into three parts, including coarse-grained financial opinion mining, fine-grained financial opinion mining, and possible research directions. This tutorial starts by introducing the components in a financial opinion proposed in our research agenda (Chen et al., 2021b) and summarizes their related studies. We also highlight the task of mining customers' opinions toward financial services in the FinTech industry, and compare them with usual opinions. Several potential research questions will be addressed. The audiences of this tutorial will gain an overview of financial opinion mining and figure out their research directions.

Tutorial Outline
We will cover the following topics based on recent works published in representative conferences and workshops. Both technical details and the application scenarios will be introduced. The contrast of financial opinion mining with general opinion mining will also be discussed. The characteristics of different kinds of financial documents will be listed.

Coarse-grained Financial Opinion Mining
The topic of the first session gives the overview of financial opinion mining, including the investor's opinion and the customer's opinion. We start with sentiment analysis in the financial domain. The comparison between the general sentiment and the market sentiment will also be discussed ( This session also covers the sentiment analysis of financial narratives from different resources, including formal documents such as financial statements and professional analyst's reports and informal documents such as blogs and social media platforms. The overview of applications on stock movement prediction and volatility forecasting will also be presented.

Fine-grained Financial Opinion Mining
The second session will focus on the fine-grained financial opinion mining, which is the recent trend in this field and also the research interest of the presenters. This session will start by the discussion of the aspect analysis of financial narratives (Maia et al., 2018;Chen et al., 2019a). The numeral in the textual data (Naik et al., 2019;Wallace et al., 2019;Chen et al., 2018Chen et al., , 2019aChen et al., , 2020c and the numeracy of the neural network models (Spithourakis and Riedel, 2018;Chen et al., 2019b) attract lots of attentions recently. In the financial narrative, the proportion of numerals are higher than that of other domains' documents. Without numerals, more important information will be missed. Thus, we summarize the related works for understanding the numerals in financial documents and provide a systematic analysis on these studies. The linguistic features of different kinds of financial documents will also be discussed (Keith and Stent, 2019;Zong et al., 2020), which can provide insights for the future works on feature engineering. The results of cross-document inference and comparison are also included (Chen et al., 2018;Keith and Stent, 2019).

Possible Research Directions
In the last session, we will discuss four possible research directions for future works (Chen et al., 2020a), including (1) argument mining in finance, (2) opinion quality evaluation, (3) implicit influence inference, and (4) opinion tracking in time series. We will link the proposed directions with the latest progress of NLP. For example, when introducing the ideas of argument mining in finance, we will provide a brief overview of current development on argument mining (Cabrio and Villata, 2018; Lawrence and Reed, 2019), and further present some instances for discussing the relation between current works and the proposed directions in financial opinion mining (Chen et al., 2020c). When discussing opinion quality evaluation, we will cover the studies of online review quality evaluation (Eirinaki et al., 2012;Wei et al., 2016;Ocampo Diaz and Ng, 2018), and show the difference between dealing with online reviews and dealing with financial opinions. The audience will be inspired by this tutorial and find an interesting research direction for their work. With the discussion on the possible research directions, many novel ideas will be figured out during this tutorial.

Recommended Small Reading List
We recommend the audiences to read the following papers, which will be discussed in the tutorial.