Inducing stock market lexicons from disparate Chinese texts

Abstract

Purpose The purpose of this paper is to propose a methodology to construct a stock market sentiment lexicon by incorporating domain-specific knowledge extracted from diverse Chinese media outlets. Design/methodology/approach This paper presents a novel method to automatically generate financial lexicons using a unique data set that comprises news articles, analyst reports and social media. Specifically, a novel method based on keyword extraction is used to build a high-quality seed lexicon and an ensemble mechanism is developed to integrate the knowledge derived from distinct language sources. Meanwhile, two different methods, Pointwise Mutual Information and Word2vec, are applied to capture word associations. Finally, an evaluation procedure is performed to validate the effectiveness of the method compared with four traditional lexicons. Findings The experimental results from the three real-world testing data sets show that the ensemble lexicons can significantly improve sentiment classification performance compared with the four baseline lexicons, suggesting the usefulness of leveraging knowledge derived from diverse media in domain-specific lexicon generation and corresponding sentiment analysis tasks. Originality/value This work appears to be the first to construct financial sentiment lexicons from over 2m posts and headlines collected from more than one language source. Furthermore, the authors believe that the data set established in this study is one of the largest corpora used for Chinese stock market lexicon acquisition. This work is valuable to extract collective sentiment from multiple media sources and provide decision-making support for stock market participants.

How to identify influential content: Predicting retweets in online financial community

“…In addition, another interesting finding is that for any chosen classifier, the performance of predicting whether a tweet will be retweeted or not is relatively better than that of predicting retweet volume in terms of all result indicators. This result echoes the findings of Zhao et al . (2020).…”

“…Using the five sets of features mentioned above, RF shows the best prediction performance for the two different dependent variables, which is consistent with the results reported in previous studies on users' retweet prediction (Sharma and Gupta, 2022). Furthermore, the superiority of the random forest algorithm has also been verified in other fields, such as content donation prediction (Zhao et al ., 2020), online reviews helpfulness prediction (Lee et al ., 2018), crash injury severity prediction (Santos et al ., 2022) and even energy consumption prediction (Ding et al ., 2021). In addition, another interesting finding is that for any chosen classifier, the performance of predicting whether a tweet will be retweeted or not is relatively better than that of predicting retweet volume in terms of all result indicators.…”

“…First, in terms of the author's basic features, the gender of an author was considered because prior research has shown the role of gender in users' donation behavior (Zhao et al ., 2020). The gender in this work could be female, male, confidential and default null value.…”

“…These features were extracted before data analysis, and are expected to influence users' decisions according to previous research on other topics, such as information persuasion (Fishbein and Ajzen, 1981), review helpfulness (Hong et al ., 2017), and user engagement (Chen et al ., 2020). More importantly, Jieba, a popular language analysis tool that focuses on Chinese sentence segmentation (Zhao et al ., 2020), was utilized for word segmentation and part-of-speech (POS) tagging. Then the number of nouns, verbs, adjectives and modal particles were incorporated as features of POS tags.…”

Tracing the technological trajectory of coal slurry pipeline transportation technology: An HMM-based topic modeling approach

“…Word2vec provides a modeling method that extracts feature vectors of words from their contexts to express deep semantic information about the words (Mikolov et al, 2013). Word2vec is also an effective tool to obtain semantic similarity (Zhao et al, 2020). Therefore, to better quantitatively analyze technology trends, we combine the LDA and Word2vec to capture technical information about potential topics in patent texts at the semantic level.…”