A Generative Approach for Financial Causality Extraction

Nayak, Tapas; Sharma, Soumya; Butala, Yash; Dasgupta, Koustuv; Goyal, Pawan; Ganguly, Niloy

doi:10.1145/3487553.3524633

Cited by 7 publications

(3 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table 1. Experimental results on the MedCasual P(%) R(%) F(%) BERT-softmax,2018 [12] 75.66 75.11 75.38 GNN,2020 [13] 51.90 72.54 60.51 BiLSTM-CRF,2021 [14] 68.78 63.23 65.89 GMTL,2022 [15] 82.79 79.57 81.15 CEPN,2022 [16] 83 [12] 79.63 75.31 77.41 GNN,2020 [13] 56.36 66.89 61.18 BiLSTM-CRF,2021 [14] 71.20 68.34 69.74 GMTL,2022 [15] 80.49 79.59 80.04 CEPN,2022 [16] 89…”

Section: Baseline Model Comparison Experimentsmentioning

confidence: 99%

Causal relationship recognition algorithm based on external semantic and contextual structural features

Li,

Shuang

2024

ACETR

View full text Add to dashboard Cite

Identifying the causal relationship of events plays an important role in determining the development of known events and evaluating the possible outcomes of different decisions. At present, neural network models are widely used to identify the relationships between events. Based on obtaining events, researchers distinguish the relationships between events by mining their semantics. However, due to the complexity of events and the dynamic changes in relationships between events, models often cannot fully meet the needs of accurately identifying causal relationships by only learning simple event descriptions in sentences; Moreover, focusing too much on the events themselves often leads to neglecting the structural features of statements and neglecting the impact of specific structural patterns on the relationships between events. In this article, we propose an event masking algorithm that combines external semantics to address the aforementioned issues. In this algorithm, external semantics are first introduced into the statement to enrich the information behind the event, allowing the model to mine the deep connections between events through a wider range of background knowledge; Then, the event masking module is used to enhance the model's extraction of sentence structured features, mining specific contextual representations that are unrelated to the event. The results show that compared to existing neural network algorithms, the algorithm proposed in this paper improves the F1 value of predictions on publicly available datasets by more than 4%.

show abstract

Section: Baseline Model Comparison Experimentsmentioning

confidence: 99%

Causal relationship recognition algorithm based on external semantic and contextual structural features

Li,

Shuang

2024

ACETR

View full text Add to dashboard Cite

show abstract

“…They used the SemEval-2010 task-8 dataset for the entire training. [67], present a generative technique for causality extraction using pointer networks and an encoder-decoder framework. They used financial domain and FinCausal for experiments and they achieved very competitive performance on this dataset.…”

Section: Introductionmentioning

confidence: 99%

Towards Improving Causality Mining using BERT with Multi-level Feature Networks

2022

KSII TIIS

View full text Add to dashboard Cite

Causality mining in NLP is a significant area of interest, which benefits in many daily life applications, including decision making, business risk management, question answering, future event prediction, scenario generation, and information retrieval. Mining those causalities was a challenging and open problem for the prior non-statistical and statistical techniques using web sources that required hand-crafted linguistics patterns for feature engineering, which were subject to domain knowledge and required much human effort. Those studies overlooked implicit, ambiguous, and heterogeneous causality and focused on explicit causality mining. In contrast to statistical and non-statistical approaches, we present Bidirectional Encoder Representations from Transformers (BERT) integrated with Multilevel Feature Networks (MFN) for causality recognition, called BERT+MFN for causality recognition in noisy and informal web datasets without human-designed features. In our model, MFN consists of a three-column knowledge-oriented network (TC-KN), bi-LSTM, and Relation Network (RN) that mine causality information at the segment level. BERT captures semantic features at the word level. We perform experiments on Alternative Lexicalization (AltLexes) datasets. The experimental outcomes show that our model outperforms baseline causality and text mining techniques.

show abstract

“…Market data has received significant interest from the academic community, mainly focusing on time series of stock prices due to the availability of this data. Although many techniques have been developed in an attempt to infer the generative processes that result in the observed prices [3,4,5,6], due to data sharing constraints there has been less focus on the microscopic behaviors of financial systems through the study of the transactions themselves. The temporal dimension has been studied by observing the arrival times of orders in limit order books, electronic records of the outstanding orders in individual stocks, as this gives a view of the supply and demand in the market.…”

Section: Introductionmentioning

confidence: 99%

Modelling Equity Transaction Networks as Bursty Processes

Seabrook¹,

Barucca²,

Caccioli³

2022

Preprint

View full text Add to dashboard Cite

Trade executions for major stocks come in bursts of activity, which can be partly attributed to the presence of self-and mutual excitations endogenous to the system. In this paper, we study transaction reports for five FTSE 100 stocks. We model the dynamic of transactions between counterparties using both univariate and multivariate Hawkes processes, which we fit to the data using a parametric approach. We find that the frequency of transactions between counterparties increases the likelihood of them to transact in the future, and that univariate and multivariate Hawkes processes show promise as generative models able to reproduce the bursty, hub dominated systems that we observe in the real world. We further show that Hawkes processes perform well when used to model buys and sells through a central clearing counterparty when considered as a bivariate process, but not when these are modelled as individual univariate processes, indicating that mutual excitation between buys and sells is present in these markets.

show abstract

A Generative Approach for Financial Causality Extraction

Cited by 7 publications

References 1 publication

Causal relationship recognition algorithm based on external semantic and contextual structural features

Causal relationship recognition algorithm based on external semantic and contextual structural features

Towards Improving Causality Mining using BERT with Multi-level Feature Networks

Modelling Equity Transaction Networks as Bursty Processes

Contact Info

Product

Resources

About