Paper: arXiv 2403.00774
Authors: Vasilii Chsherbakov, Ilia Karpov
Abstract
Inflation is one of the most important macroeconomic indicators that have a great impact on the population of any country and region. Inflation is influenced by range of factors, one of which is inflation expectations. Many central banks take this factor into consideration while implementing monetary policy within the inflation targeting regime. Nowadays, a lot of people are active users of the Internet, especially social networks. There is a hypothesis that people search, read, and discuss mainly only those issues that are of particular interest to them. It is logical to assume that the dynamics of prices may also be in the focus of user discussions. So, such discussions could be regarded as an alternative source of more rapid information about inflation expectations. This study is based on unstructured data from Vkontakte social network to analyze upward and downward inflationary trends (on the example of the Omsk region). The sample of more than 8.5 million posts was collected between January 2010 and May 2022. The authors used BERT neural networks to solve the problem. These models demonstrated better results than the benchmarks (e.g., logistic regression, decision tree classifier, etc.). It makes possible to define pro-inflationary and disinflationary types of keywords in different contexts and get their visualization with SHAP method. This analysis provides additional operational information about inflationary processes at the regional level The proposed approach can be scaled for other regions. At the same time the limitation of the work is the time and power costs for the initial training of similar models for all regions of Russia.
Complexity vs Empirical Score
- Math Complexity: 3.5/10
- Empirical Rigor: 7.0/10
- Quadrant: Street Traders — empirical and applied, lower math complexity
Why this score: The paper primarily applies established NLP techniques (BERT, SHAP) rather than developing new mathematical theory, but it demonstrates strong empirical rigor with a large-scale dataset (8.5M posts), robust methodology, and clear operational relevance for real-world forecasting.
Research Flowchart
flowchart TD Start["Research Goal: Analyze inflation trends via social media data"] --> Input["Data: 8.5M+ posts from Vkontakte (Jan 2010 - May 2022)"] Input --> NLP["NLP Processing: BERT neural networks trained for classification"] NLP --> Methodology["Methodology: Identify pro-inflationary vs. disinflationary keywords"] Methodology --> SHAP["SHAP Analysis: Visualize key features & context impact"] SHAP --> Outcomes["Outcomes: Operational insights on regional inflation expectations"] Outcomes --> Scale["Outcome: Scalable approach for other regions (limited by compute cost)"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes using VK social network posts and a RuBERT‑tiny based model (BERT‑inflation‑256) to infer regional inflation expectations and identify pro‑inflationary and disinflationary keywords.
Source sentence
We used RuBERT-tiny (state-of-art model) to build 4 variants of models depending on the maximum length of sentences. Based on the f1-score, as well as the time resources needed to train models, BERT-inflation-256 was chosen as the best model.
Assumes 3
- User discussions on social networks reflect price dynamics and can serve as a proxy for inflation expectations. data
Source sentence
It is logical to assume that the dynamics of prices may also be in the focus of users’ discussions.
- People discuss on the Internet mainly issues of particular interest to them. other
Source sentence
There is a hypothesis that people search, read, and discuss mainly only those issues that are of particular interest to them.
- Groups with >2 000 users and ≥20 % Omsk residents represent the Omsk region population. data
Source sentence
We applied some empirical criteria in order to determine the original Omsk group based on prevailing users from the Omsk region. So, we used only groups with more than 2 000 users and in which the share of Omsk residents was at least 20%.
Data
- VK Omsk region posts text-news public
Source sentence
The sample of more than 8.5 million posts was collected between January 2010 and May 2022.
Twitter / X
Costs
Not applicable
Finds 3
- The BERT‑inflation‑256 model achieved higher predictive performance than benchmark classifiers such as logistic regression, decision tree, random forest, and gradient boosting. predictive-accuracy
Source sentence
This model demonstrated better results than the benchmark models (logistic regression, decision tree classifier, random forest classifier, and gradient boosting classifier).
- BERT‑inflation‑256 was selected as the best model based on f1‑score and training time efficiency. performance
Source sentence
Based on the f1-score, as well as the time resources needed to train models, BERT-inflation-256 was chosen as the best model.
- SHAP analysis enables identification of pro‑inflationary and disinflationary keywords in different contexts. other
Source sentence
We applied SHAP for an interpretation of this model. It makes possible to define pro-inflationary and disinflationary types of keywords in different contexts.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 20 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.