PaperlayerView exact revision on arXiv

Source-linked AI summary

CognitiveSky: Scalable Sentiment and Narrative Analysis for Decentralized Social Media

Gaurab Chhetri, Anandi Dutta, Subasish Das

arXiv:2509.11444v1cs.CLcs.SI

TL;DR

Limited access to Twitter/X APIs has shifted research attention toward decentralized platforms, while existing analysis tools may be static or limited in functionality. CognitiveSky provides a modular framework for analyzing live Bluesky discourse with NLP labeling and dashboard summaries; its mental health case study captures nuanced emotional expressions and evolving themes, though the current system has language, rate-limit, dashboard, and data-volatility constraints.

  • Problem

    Twitter/X API restrictions limit academic and open-source access, while existing discourse-analysis tools have limits in scalability, usability, analytical depth, or support for live decentralized data.

    Source: §1 · p. 1

  • Method

    CognitiveSky is an open-source modular pipeline that ingests Bluesky discourse, applies transformer-based labeling, and produces summaries and visualizations for analysis.

    Source: §1 · pp. 1–2, §2 · p. 2

  • Results

    The mental health case study demonstrated the system’s ability to capture nuanced emotional expressions and evolving themes within online communities.

    Source: §6 · p. 6

  • Takeaways & Limitations

    Its modular design and open-source, automated pipeline can support research and monitoring across domains including disinformation, crisis detection, social movements, and policy response.

    Source: §5 · pp. 5–6

    The current pipeline relies on English-only models, faces real-time scalability constraints from rate limits and API policies, has static-dashboard limits, and encounters data volatility that challenges replicability and long-term validation.

    Source: §6 · p. 6

Abstract

from arXiv · show

The emergence of decentralized social media platforms presents new opportunities and challenges for real-time analysis of public discourse. This study introduces CognitiveSky, an open-source and scalable framework designed for sentiment, emotion, and narrative analysis on Bluesky, a federated Twitter or X.com alternative. By ingesting data through Bluesky's Application Programming Interface (API), CognitiveSky applies transformer-based models to annotate large-scale user-generated content and produces structured and analyzable outputs. These summaries drive a dynamic dashboard that visualizes evolving patterns in emotion, activity, and conversation topics. Built entirely on free-tier infrastructure, CognitiveSky achieves both low operational cost and high accessibility. While demonstrated here for monitoring mental health discourse, its modular design enables applications across domains such as disinformation detection, crisis response, and civic sentiment analysis. By bridging large language models with decentralized networks, CognitiveSky offers a transparent, extensible tool for computational social science in an era of shifting digital ecosystems.

1. Introduction

Twitter/X has been a key source for studying public discourse, but API restrictions have limited academic access as attention shifts toward more open decentralized platforms such as Bluesky. CognitiveSky extends interactive discourse analysis to live Bluesky data through a modular framework for labeling and summarizing posts.

1. Introduction

  • API restrictions have limited academic and open-source access to Twitter/X data as researchers turn toward more open, decentralized platforms such as Bluesky.Social media platforms are described as central to public discourse, and Twitter historically served as a core research dataset.Source: §1 · p. 1
  • Earlier tools offer partial narrative-analysis capabilities, while TwiXplorer combines sentiment analysis, topic modeling, and visualization but remains limited to static Twitter/X archives.CognitiveSky is presented as extending that paradigm to real-time Bluesky discourse.Source: §1 · p. 1
  • CognitiveSky continuously ingests live Bluesky data, applies transformer-based sentiment and emotion detection, and produces structured JSON summaries for interactive visualization.The framework extends earlier interactive analysis into real-time decentralized discourse analysis.Source: §1 · p. 1
  • Unlike tools built around static archives, CognitiveSky is designed to be lightweight, low-cost, scalable, and adaptable across domains including mental health, disinformation, civic engagement, and crisis response.Table 1 compares CognitiveSky and TwiXplorer in architectural and functional capabilities.Source: §1 · p. 1, Table 1 · p. 2
  • The paper proceeds to describe CognitiveSky’s architecture, NLP pipeline, summarization engine, dashboard, and broader applicability.Source: §1 · p. 2

2. System Overview

CognitiveSky is an open-source pipeline that ingests public Bluesky posts, labels them with transformer models and topic clustering, then aggregates results into JSON snapshots for dashboard exploration. Its three loosely coupled layers support real-time collection, scheduled annotation, and interactive analysis while presenting only aggregate data.

2. System Overview

  • The system separates ingestion, labeling, and summarization into three loosely coupled layers for real-time streaming, daily summaries, and long-term trend analysis.It is an open-source modular pipeline for discourse data from Bluesky.Source: §2 · p. 2

2.1. Ingestion and Preprocessing

  • A 24/7 Node.js service listens to public Bluesky Firehose posts over WebSocket, filters them with case-insensitive keyword patterns, and stages records in Supabase for labeling.Text cleaning and validation are deferred to annotation; buffered records are batch-inserted every five seconds or after 200 posts, with retry and graceful-shutdown handling.Source: §2.1 · p. 2

2.2. Data Labeling

  • A daily GitHub Actions workflow processes batches of up to 64 posts with transformer-based sentiment and emotion models, then uses TF-IDF and MiniBatch NMF to assign interpretable topic clusters.Sentiment labels are positive, neutral, or negative; emotion detection uses a DistilRoBERTa model trained on GoEmotions, and representative topic keywords are retained for visualization.Source: §2.2 · pp. 2–3

2.3. Real-time Interactive Dashboard

  • Daily and weekly JSON snapshots aggregate temporal counts, sentiment shifts, emotional variation, topic frequencies and keywords, hashtag and emoji co-occurrences, and global metrics.The default snapshot configuration covers the past seven days, with longer windows configurable.Source: §2.3 · p. 3
  • The static-snapshot dashboard offers an overview plus views for sentiment, emotion, topics, hashtags, emojis, and timeline comparisons; Figure 2 shows word clouds, bar charts, and summary metrics.Static JSON supports low-latency access without live database queries. The interface exposes aggregate analysis rather than individual post text, metadata, or identifiers.Source: §2.3 · pp. 3–4, Figure 2 · p. 4
  • The system processes only publicly accessible, non-reply posts, references accounts by Bluesky DIDs, and does not expose individual-level posts through its interface or downloadable datasets.Text is retained and processed for aggregate statistical and thematic analysis.Source: §2.3 · p. 4

3. Technology Optimizations

CognitiveSky is built as a modular, reproducible pipeline using open-source and free-tier infrastructure. Its collection, labeling, and visualization work is divided across JavaScript/TypeScript, Python, and GitHub Actions.

3. Technology Optimizations

  • The system is designed as a lightweight, modular, reproducible pipeline whose layers operate autonomously using open-source and free-tier infrastructure.The technology-stack table is described as covering functional roles and rationale.Source: §3 · p. 4, Table 2 · p. 5
  • JavaScript and TypeScript support data collection and visualization, while a Python pipeline handles data labeling and processing.Source: §3 · p. 4
  • GitHub Actions serves as the CI layer for automated daily runs of the Python-based pipeline.The passage links reproducibility with community-driven development and open social data science.Source: §3 · pp. 4–5

4. Use Case: Mental Health Narratives

CognitiveSky is first deployed for mental health discourse on Bluesky, where emotionally intense discussion and subtle contextual cues make the domain a testbed for its labeling and visualization pipeline. The use case focuses on what people discuss, how they feel, when they engage, and the language or cultural markers they use.

4. Use Case: Mental Health Narratives

  • The initial deployment focuses on mental health, motivated by concern about online anxiety, depression, and emotional distress, particularly among younger populations adopting platforms such as Bluesky.Source: §4 · p. 5
  • Mental health discussion offers a testbed for the pipeline because it can involve heightened emotional intensity, subtle contextual cues, and implicit signals of distress or calls for help.The analysis captures discussion topics, feelings, engagement timing, and linguistic or cultural markers of emotional states.Source: §4 · p. 5
  • The use case continuously monitors the Bluesky Firehose for public posts containing mental-health-related terms.Source: §4 · p. 5

5. Replicable Framework and Other Use Cases

CognitiveSky is presented as adaptable beyond mental-health discourse, with potential applications in disinformation and crisis monitoring, civic analysis, and interdisciplinary research.

5. Replicable Framework and Other Use Cases

  • Its modular design supports applications beyond mental-health narratives, including information security, crisis detection, social movement tracking, and academic inquiry.The framework extracts structured insights from unstructured, real-time, decentralized social media data.Source: §5 · p. 5
  • For disinformation and emergencies, CognitiveSky combines clustering, sentiment, and emotion analysis to monitor semantically similar posts, affective patterns, and evolving topics.The passage identifies spikes in fear, anger, or outrage as examples of affective patterns in disinformation campaigns.Source: §5 · p. 5
  • In civic settings, it supports fine-grained analysis of public sentiment and emotional tone during movements, elections, and legislative debates.Optional metadata such as geolocation supports regional and demographic comparisons.Source: §5 · p. 6
  • Its modular pipeline and GitHub Actions automation lower infrastructure barriers for longitudinal discourse, semantic-evolution, and policy-response research.The passage describes intended use by social scientists, linguists, and public health researchers.Source: §5 · p. 6

6. Conclusions

CognitiveSky is an open-source, real-time system for analyzing discourse trends on Bluesky, with a mental-health case study illustrating emotional and thematic analysis. The conclusion notes limits in language coverage, real-time scalability, dashboard interactivity, and long-term validation.

6. Conclusions

  • The mental-health case study illustrates the system’s ability to capture nuanced emotional expressions and evolving community themes.CognitiveSky is described as an open-source, real-time system for detecting, labeling, and visualizing Bluesky discourse trends.Source: §6 · p. 6
  • English-only models, API policies and rate limits, static dashboards, and network data volatility constrain language coverage, real-time scalability, interactivity, and long-term validation.The conclusion also identifies free-tier deployment, serverless infrastructure, and modular pipelines as architectural choices that avoid costly infrastructure.Source: §6 · p. 6
  • Future work includes multilingual and cross-cultural expansion, improved narrative-shift detection and topic tracking, longitudinal user analysis, and more interactive dashboards.The conclusion frames these developments as ways the system could evolve across a broader range of decentralized ecosystems.Source: §6 · p. 6
Loading 2509.11444v1…