TEXTUAL: International Workshop on Text Analytics and Knowledge Extraction From Natural Language Software Artifacts

About TEXTUAL 2027

Natural language is the connective tissue of modern software engineering. While source code defines what a system executes, millions of software repositories store the vital human intent, architectural rationale, and collaborative decision-making behind why and how software evolves. Critical domain knowledge remains locked inside unstructured and semi-structured text across the entire software development lifecycle (SDLC), ranging from traditional Requirements specifications, User stories, Architectural Decision Records (ADRs), Bug reports, Test case specifications, App Descriptions, Developer communications (GitHub discussions, Slack, pull request code reviews), and post-release user feedback, App reviews, to emerging artifacts like system prompts, Agentic skills (SKILL.md), and Markdown documentation.

TEXTUAL brings together researchers and practitioners at the intersection of Software Engineering (SE), Mining Software Repositories (MSR), Natural Language Processing (NLP), and Information Retrieval (IR). The workshop provides a dedicated forum to advance data-driven methodologies, repository-scale empirical evaluations, traditional text mining techniques (such as topic modeling, classification, and IR-based traceability), as well as domain-adapted LLMs, prompt and agent analytics, and open-access benchmark datasets designed to analyze, link, and extract actionable software intelligence from technical text.

Topics of Interest

We invite original contributions addressing quantitative, qualitative, empirical, statistical, and AI-driven approaches to software text analytics across the SDLC. Topics of interest include, but are not limited to:

Repository Mining & Text-Centric MSRLarge-scale mining of text repositories (GitHub, Jira, Stack Overflow, Slack), .md documentation ecosystems, commit logs, and process metrics extraction.
Requirements Engineering & Specification AnalyticsAmbiguity detection, quality assessment, quality concerns extraction, inconsistency resolution, and structural model extraction from SRSs, User Stories, and feature specs.
Architecture Rationale & Design Knowledge MiningCapturing, classifying, and synthesizing design rationale, quality concerns, architectural constraints, and technical debt from ADRs, RFCs, and architecture wikis.
Developer Communication & Social AnalyticsMining developer interactions to analyze concerns, team dynamics, sentiment, tone, toxicity, onboarding friction, and collaborative decision-making.
Maintenance & Post-Release Feedback MiningAutomated triage, duplicate detection, severity prediction for bug reports, and mining feature requests/issues from App Store reviews.
Information Retrieval & TraceabilityLexical, sparse, dense, and hybrid IR models including BM25, VSM, semantic embeddings (S-BERT), bi-encoders, late-interaction architectures (ColBERT), and neural rerankers for automated traceability recovery across heterogeneous artifacts (e.g., SRSs to test suites, ADRs to bug reports).
Topic Modeling & Semantic StructuringLDA, BERTopic, structural topic modeling, and clustering for software documentation, developer discussions, and code comments.
Prompts, Agentic Skills & LLM Instruction MiningEmpirical analysis, quality metrics, static/dynamic evaluation, and versioning of system prompts, agentic skills (SKILL.md), and LLM tool definitions.
LLMs, Datasets & Domain AdaptationFine-tuning, RAG pipelines, hallucination/bias mitigation for SE dialects, annotated open-access corpora, and benchmark evaluation standards.

Submission Categories

All papers must be written in English, formatted according to the official IEEE Conference Proceedings guidelines (two-column layout), and submitted electronically via the workshop submission portal.

We invite the following paper categories:

Full Research PapersUP TO 8 PAGESMature research contributions, thorough empirical investigations, or validated technical methodologies.
Short / Vision PapersUP TO 4 PAGESWork-in-progress, novel research directions, preliminary empirical results, or visionary challenge proposals.
Tool & Dataset PapersUP TO 4 PAGESDemonstrations of automated text processing pipelines, mining tools, or annotated open-access software text corpora.

Submission Site Coming Soon

Workshop Format

TEXTUAL is planned as a full-day interactive workshop. The program will combine invited and peer-reviewed content with structured discussions to encourage community building around text analytics for software engineering.

Opening and workshop framing
Invited keynote talk
Presentations of accepted research, vision, tool, and dataset papers
Panel on text analytics and LLMs in industrial software practice
Structured breakout discussion on dataset, benchmark, and evaluation gaps
Closing synthesis to identify community priorities and collaboration opportunities