Purpose: This study examines whether domain-specific public mood is associated with stock-market dynamics and whether these relationships vary across communication domains and market conditions. It distinguishes public mood from investor sentiment and investigates whether heterogeneous sources of public communication exhibit different relationships with market behaviour. Design: The study analyses 610,422 posts published by 176 curated X accounts between January 2022 and December 2023, covering Politics and Government, Economy and Finance, and Media and Society. Posts are classified using fine-tuned Turkish transformer models under three domain-specific and one pooled regime. Public mood measures are constructed at daily, weekly, and monthly frequencies and examined alongside BIST100 and BIST30 market measures using correlation, Granger causality, vector autoregression, and impulse response analyses across the full period and selected market conditions. Findings: Public mood is not associated with the direction of stock-market returns but is associated with the magnitude of price movements, particularly for Media and Society and pooled communication. These relationships become stronger at longer aggregation frequencies. Predictive relationships are concentrated in Economy and Finance communication, while their magnitude and direction vary across market conditions, particularly during the 2023 election period. The pooled measure largely reflects the most active communication domain. Originality: The study contributes to behavioral-finance research by incorporating communication - domain heterogeneity into the analysis of public mood and market dynamics. It also demonstrates how aggregating heterogeneous sources can obscure domain-specific relationships between public communication and financial markets.
Figures & tables
Figure 1: Methodological outline of the study.
Dataset
Total
Train
Valid.
Test
Positive
Negative
Neutral
Politics and Government
1,500
1,000
200
300
959
460
81
Economy and Finance
1,506
1,006
200
300
705
652
149
Media and Society
1,502
1,002
200
300
764
599
139
Combined
4,508
3,008
600
900
2,428
1,711
369
Table 1: Labelled dataset sizes and class distribution.
Labelling regime
Selected model
Accuracy
Weighted F1
Macro F1
Politics and Government
BERTurk
93.0
92.9
86.9
Economy and Finance
ConvBERTurk
86.0
85.6
79.7
Media and Society
BERTurk
86.3
86.6
80.4
Combined
ConvBERTurk
88.6
88.1
81.7
Table 2: Best single run of the selected classification model for each labelling regime.
BIST100
BIST30
Series
ADF
p
Class
ADF
p
Class
Open
− 0.635
0.8629
NS
− 0.632
0.8635
NS
Close
− 0.719
0.8417
NS
− 0.711
0.8438
NS
Volume
− 2.670
0.0794
NS
− 3.430
0.0100
S
Return
− 9.755
0.0000
S
− 10.000
0.0000
S
Intraday change %
− 10.033
0.0000
S
− 10.220
0.0000
S
Table 3: Augmented Dickey–Fuller tests, daily series, full period.
BIST100
BIST30
Variable
Pol.
Econ.
Media
Comb.
Pol.
Econ.
Media
Comb.
Close ‡
0.093 *
0.085 †
0.328 ***
0.347 ***
0.098 *
0.076 †
0.338 ***
0.355 ***
Volume §
− 0.095 *
− 0.027
0.205 ***
0.079 †
− 0.081 †
− 0.033
0.139 **
0.038
Return
− 0.007
0.004
0.044
0.047
− 0.014
− 0.005
0.040
0.045
Intraday change %
0.010
− 0.020
0.016
0.020
0.005
− 0.029
0.019
0.023
Range §
0.029
− 0.010
0.257 ***
0.216 ***
0.039
− 0.024
0.289 ***
0.239 ***
Table 4: Daily correlations between public mood and market variables, full period.
BIST100
BIST30
Variable
Pol.
Econ.
Media
Comb.
Pol.
Econ.
Media
Comb.
Range, daily
0.029
− 0.010
0.257 ***
0.216 ***
0.039
− 0.024
0.289 ***
0.239 ***
Range, weekly
0.087
0.128
0.363 ***
0.372 ***
0.098
0.100
0.408 ***
0.402 ***
Range, monthly
0.214
0.214
0.482 *
0.523 **
0.233
0.162
0.531 **
0.558 **
Return, daily
− 0.007
0.004
0.044
0.047
− 0.014
− 0.005
0.040
0.045
Return, weekly
0.160
− 0.046
0.108
0.097
0.173 †
0.002
0.097
0.111
Table 5: Public mood and market correlations across aggregation frequencies.
Daily
Weekly
Index
Direction
Pol.
Econ.
Media
Comb.
Pol.
Econ.
Media
Comb.
BIST100
Mood → Market
1/28
11/28
1/28
2/28
0/28
7/28
0/28
0/28
BIST100
Market → Mood
1/28
5/28
2/28
3/28
0/28
0/28
1/28
1/28
BIST30
Mood → Market
0/28
9/28
0/28
0/28
1/28
4/28
0/28
0/28
BIST30
Market → Mood
1/28
3/28
0/28
0/28
0/28
0/28
1/28
1/28
Table 6: Significant Granger causality tests, full period.
Sub-period
Tests per direction
Mood → Market
Market → Mood
Panel A: BIST100
2022 baseline
112
6
0
2023 Q1 earthquake
112
4
7
2023 Q2 election
112
4
43
2023 H2 IPO boom
112
3
10
Panel B: BIST30
Table 7: Significant daily Granger causality tests by sub-period and direction.
BIST100
BIST30
Variable
Pol.
Econ.
Media
Comb.
Pol.
Econ.
Media
Comb.
Daily return
2022 baseline
0.054
0.100
0.105 †
0.177 **
0.048
0.098
0.096
0.167 **
2023 Q1 earthquake
0.070
0.013
− 0.045
0.067
0.055
− 0.003
− 0.046
0.064
2023 Q2 election
− 0.228 †
− 0.164
− 0.095
− 0.281 *
− 0.191
− 0.178
− 0.061
− 0.247 †
2023 H2 IPO boom
− 0.019
− 0.035
0.149 †
0.118
− 0.056
− 0.056
0.121
0.088
Table 8: Daily correlations by sub-period.
BIST100
BIST30
Category
Accounts
Return
Range %
Vol 21d
Return
Range %
Vol 21d
Politics and Government
70
1 (1.4)
3 (4.3)
17 (24.3)
1 (1.4)
5 (7.1)
15 (21.4)
Economy and Finance
67
11 (16.4)
12 (17.9)
8 (11.9)
11 (16.4)
13 (19.4)
8 (11.9)
Media and Society
64
3 (4.7)
7 (10.9)
15 (23.4)
3 (4.7)
6 (9.4)
14 (21.9)
All accounts pooled
169
12 (7.1)
22 (13.0)
33 (19.5)
11 (6.5)
21 (12.4)
33 (19.5)
Table 9: Accounts correlating significantly with selected market variables, by category.
Appendix figures & tables1 asset
Supplementary material from the paper’s appendix.
Appendix
Model
Accuracy
Weighted F1
Precision
Recall
BERTurk
93.0
92.9
93.0
93.0
ConvBERTurk
92.3
92.1
92.5
92.3
DistilBERTurk
89.0
88.7
89.0
89.0
TurkishBERTweet
88.7
88.3
88.4
88.7
mBERT
87.0
86.8
87.0
87.0
Gemma-3-1B
87.3
86.6
87.0
87.3
Appendix
Table A.1: Best single run of each of the seven fine-tuned models on the Politics and Government dataset, sorted by weighted F1. All values are percentages.
Emojis are widely used in online financial communication, but it is unclear whether they provide transferable sentiment signals across languages, platforms, and asset communities. This study examines the extent to which emoji usage, semantics, and sentiment polarity remain stable across financial communities, and how these layers influence zero-shot sentiment transfer. Using large corpora of Twitter and StockTwits posts in four languages, we measure cross-community divergence and evaluate sentiment models trained under emoji-only, text-only, and text+emoji inputs. We find that emoji frequencies differ across communities, especially across languages, but their semantics and sentiment polarity are largely stable. Cross-asset transferability shows minimal degradation, while cross-language transfer remains the most challenging. Including emojis consistently reduces transfer gaps relative to text-only models. These results indicate that financial communication exhibits a partially shared ``emoji code,'' and that emojis provide compact, language-independent sentiment cues that improve model generalization across markets and platforms.
Ahmed Mahrous, Roberto Di Pietro
King Abdullah University of Science and Technology (KAUST)
This paper explores the use of emojis in financial sentiment analysis, focusing on the social media platform StockTwits. Emojis, increasingly prevalent in digital communication, have potential as compact indicators of investor sentiment, which can be critical for predicting market trends. Our study examines whether emojis alone can serve as reliable proxies for financial sentiment and how they compare with traditional text-based analysis. We conduct a series of experiments using logistic regression and transformer models. We further analyze the performance, computational efficiency, and data requirements of emoji-based versus text-based sentiment classification. Using a balanced dataset of about 528,000 emoji-containing StockTwits posts, we find that emoji-only models achieve F1 approximately 0.75, lower than text-emoji combined models, which achieve F1 approximately 0.88, but with far lower computational cost. This is a useful feature in time-sensitive settings such as high-frequency trading. Furthermore, certain emojis and emoji pairs exhibit strong predictive power for market sentiment, demonstrating over 90 percent accuracy in predicting bullish or bearish trends. Finally, our research reveals large statistical differences in emoji usage between financial and general social media contexts, stressing the need for domain-specific sentiment analysis models.
Ahmed Mahrous, Roberto Di Pietro
King Abdullah University of Science and Technology (KAUST), Thuwal, Saudi Arabia
While sentiment analysis is the staple of financial NLP, capturing the nuances of 'why' behind that sentiment remains a challenge. There have been attempts to address this by analysing investor emotions alongside sentiment; however, this does not provide the additional granularity required to understand the target of the emotion/sentiment. We address this by augmenting the StockEmotions dataset with semantically structured opinion graphs, which provide granular semantic depth to the existing sentiment and emotion labels. Using a declarative LLM pipeline, we augment the StockEmotions dataset with opinion graphs for each sentence, derived from 10,000 comments collected from StockTwits. In addition, we study the effect of introducing opinion semantics on baseline classifiers using Graph Neural Networks (GNNs). Our analysis demonstrates that incorporating opinion semantics improves classification performance across different emotional spectrums
Gaurav Negi, Paul Buitelaar
Data Science Institute · University Of Galway · Galway, Ireland