AI Qualitative Analysis of UNGA Speeches

Evidano is an AI-powered qualitative data analysis platform that helps researchers analyze interviews, open-ended surveys, and documents. According to PLOS One, Lasekan et al. analyzed a 50, 644-token corpus of 21 African leaders' English UNGA-2025 speeches to map grammatical patterns onto UNESCO sustainability competencies. The primary keyword for this post is "AI qualitative analysis of UNGA speeches, " aimed at qualitative researchers and curriculum designers who want to apply AI methods to discourse-driven pedagogy. The payoff: concrete measures and a reproducible task model (GSIC) that AI-assisted qualitative workflows can operationalize in research and classroom practice.

Key Takeaways

According to PLOS One, a 50, 644-token corpus of 21 African leaders' UNGA-2025 English speeches (September–October 2025) shows that sustainability meaning is built mainly through vocabulary and discourse structure and can be operationalized for instruction via GSIC.

50, 644 tokens, 21 speeches, 1, 956 sentences and 1, 515 paragraphs were analyzed in September–October 2025, per PLOS One (published August 11, 2026).
The PLOS One study found 48.6% of sustainability expression at the lexical level and 26.5% at the discourse level in August 2026 reporting.
The authors report that over 72% of modal verb occurrences appeared in SDG-tagged contexts and that passive constructions appeared in 75.6% of SDG contexts, linking grammar to normative and systems-thinking competencies.
Representative quotes from the corpus include: "The UN must now focus on supporting the building of economies of the world to address issues of poverty and the global financial crisis" (UNGA 2025, Eswatini) and "It is essential that we address the climate crisis as the greatest existential threat of our time" (UNGA 2025, Angola).

What happened and how the study worked

The PLOS One study used a mixed quantitative–qualitative, corpus-informed approach to analyze grammar in sustainability discourse from UNGA-2025 speeches.

According to PLOS One, researchers compiled 21 English-language speeches delivered by African leaders during the 80th UNGA session (September–October 2025), producing a corpus of 50, 644 tokens for multi-level linguistic analysis.

The authors describe using a 327-item SDG lexicon for automatic SDG tagging, spaCy for tokenization and dependency parsing, AntConc for lexical concordances, and coder validation with Cohen's kappa values (e.g., κ =.81–.92) as reported in PLOS One.

The PLOS One team mapped lexical, morphological, syntactic, pragmatic, and discourse features to UNESCO competencies (normative, systems-thinking, interpersonal, anticipatory, strategic) and then derived the Grammar-for-Sustainability Instructional Cycle (GSIC) as a five-phase task model.

Findings snapshot

Date	Metric	Value	Implication
Sept–Oct 2025	Speeches analyzed	21 official UNGA-2025 speeches	Authentic diplomat discourse, useful for advanced EFL/ESD materials
Corpus (reported Aug 11, 2026)	Token count	50, 644 tokens	Sufficient scale for concordance-driven classroom examples
Analysis (PLOS One)	Lexical vs discourse share	48.6% lexical, 26.5% discourse	Lexico-discursive strategies are primary carriers of sustainability meaning
Syntactic findings	Modal and conditional SDG alignment	72%+ of modal verbs in SDG contexts; 72.4% conditional forms in SDG contexts	Modality and conditionals encode obligation and systems reasoning
Top SDGs (lexical frequency)	Governance & partnerships	SDG 16 ≈ 39%, SDG 17 ≈ 43.5%	Africa frames sustainability around governance, partnerships, development

Implications for qualitative and corpus researchers

The PLOS One mapping shows that grammatical features can be valid coding categories for qualitative discourse analysis, not only surface tokens.

According to PLOS One, lexical and discourse-level features accounted for 48.6% and 26.5% respectively, indicating that keyword-in-context and argument-structure coding will capture most sustainability meaning.

The PLOS One report that modal verbs and conditionals align with specific competencies (normative and systems-thinking) implies that tagging syntactic patterns (modals, passives, if–then clauses) can produce analytically rich segments for thematic coding.

Researchers using AI-enabled qualitative workflows should therefore: (1) include multi-level feature extraction (lexical, syntactic, discourse), (2) validate automated tagging against human-coded samples as PLOS One did, and (3) design coding schemes that map linguistic forms to functional competencies for cross-study comparability.

How Evidano helps researchers apply these methods

Problem: Multilevel linguistic extraction is slow and error-prone

Solution: Evidano performs automated ingestion, lemmatization, and syntactic tagging so teams can extract lexical, modal, conditional, and discourse patterns at corpus scale.

Evidano integrates transcription and parsing capabilities to prepare spoken or written speech transcripts and supports exportable concordance views for classroom GSIC tasks; see Evidano features.

Problem: Aligning linguistic features to custom competency codes is manual

Solution: Evidano supports user-definable codebooks and automated pattern matching so researchers can map modal verbs, passives, and collective pronouns to competency labels (normative, systems-thinking, interpersonal) in minutes.

Evidano also provides frequency tables and cross-segment analysis that mirror the PLOS One workflow, enabling reproducible mappings from grammar to pedagogy.

Problem: Preparing spoken-draft materials from audio recordings is time-consuming

Solution: Evidano's speech-to-text pipeline offers configurable vocabularies and PII redaction to convert UNGA audio or classroom recordings into analysis-ready transcripts.

Problem: Teachers need classroom-ready concordances and tasks

Solution: Evidano can export concordance lines, collocate lists, and example excerpts (with source attribution) to build GSIC-style notice-analyze-transform-production materials quickly, preserving provenance for citation and replication.

FAQ: AI qualitative analysis of UNGA speeches

How can AI detect grammatical features linked to sustainability competencies?

Answer: AI can tag lexical items, morphologies, syntactic patterns, and discourse structures, then map them to competency codes via rule-based or supervised labels.

According to PLOS One, the authors used a 327-item SDG lexicon, spaCy parsing, and manual validation to link features such as modals and conditionals to UNESCO competencies.

What corpus size and data quality are required to replicate PLOS One methods?

Answer: A multi-thousand-token corpus with verified transcripts is sufficient for concordance-driven insights.

The PLOS One study analyzed 50, 644 tokens from 21 speeches (September–October 2025) and reported inter-coder Cohen's kappa between.81 and.92 for validation, showing that medium-scale corpora with coder checks can deliver reliable mappings.

Can GSIC be used with low-resource classrooms or non-English corpora?

Answer: Yes, GSIC is adaptable and can use printed concordances or teacher-curated excerpts when digital tools are limited.

The PLOS One authors note that GSIC targets upper-intermediate to advanced learners and can be adapted for low-resource settings through pre-annotated worksheets; practitioners should pilot GSIC in local languages for broader inclusion.

What are the reproducibility steps researchers should follow to mirror the study?

Answer: Build an SDG lexicon, preprocess with lemmatization, tag SDG sentences, extract lexical and syntactic features, validate with human coders, and map features to competency codes.

The PLOS One article provides replication materials and supplemental CSV/JSON files for token processing and concordance outputs (see the study's Supplementary Materials linked in the article).

Conclusion & Next Steps

The PLOS One study demonstrates that a corpus-based, multi-level linguistic analysis can turn UNGA-2025 speeches into an operational pedagogy (GSIC) that teaches grammar as social action.

Qualitative researchers and curriculum designers can reproduce the approach by combining SDG lexicons, syntactic tagging, and human validation as outlined in PLOS One (published August 11, 2026).

To apply these methods at scale, use AI-enabled platforms to extract concordances, tag syntactic phenomena, and generate classroom materials efficiently; for a practical AI workflow see Evidano features and our speech-to-text tools.

Try the pipeline yourself and convert discourse analysis into classroom-ready GSIC tasks: Try Evidano for free.

Topics

AI qualitative analysis of UNGA speeches
grammar for sustainability instruction
corpus-informed qualitative research
GSIC pedagogy

Keep reading

Browse all articles