This study proposes a robust and semantically enriched framework for conversation understanding based on a composite distributed representation that incorporates both temporal adjacency and semantic proximity between utterances, enabling the visualization of key conversational connections.
Abstract
The analysis of spoken data from panel discussions, policy dialogues, and educational meetings has gained increasing importance in both academic research and professional practice. However, conventional approaches to Japanese conversation analysis have relied heavily on keyword matching or surface‑level text similarity, making it difficult to capture deeper semantic relationships, topic transitions, and latent discourse structures. In addition, Japanese natural language processing pipelines often rely on environment-sensitive morphological analyzers, which hinder reproducibility and large-scale processing. To address these limitations, this study proposes a robust and semantically enriched framework for conversation understanding based on a composite distributed representation. The proposed method integrates three layers of linguistic information: (1) contextual sentence embeddings generated by a multilingual transformer model, (2) word embeddings obtained from fastText, and (3) co‑occurrence vectors that capture lexical association patterns within the conversation. Sudachi is employed for Japanese text preprocessing to ensure stable and reproducible morphological analysis. By combining these components into a unified composite vector, the framework simultaneously represents global sentence‑level meaning and local lexical relationships. Using this representation, a directed graph is constructed that incorporates both temporal adjacency and semantic proximity between utterances, enabling the visualization of key conversational connections. To evaluate the effectiveness of the composite representation, dimensionality‑reduction algorithms are applied to examine whether semantically similar utterances naturally form coherent clusters in low‑dimensional space. The resulting clusters are assessed for consistency and interpretability, demonstrating that the proposed representation successfully captures meaningful conversational structure.
Conversational agents for low-resource languages (LRLs), such as Igbo, face major challenges, including limited annotated data, code-switching, and weak contextual coherence in multi-turn dialogue. This study proposes a multilevel-attention and hybrid-embedding framework that integrates FastText subword representations with multilingual BERT (mBERT) to improve semantic understanding and context retention. The architecture applies hierarchical attention at the word, utterance, and dialogue levels, enabling effective modeling of conversational dependencies and reducing context drift. The model was evaluated on a curated Igbo--English conversational dataset and benchmarked against long short-term memory (LSTM), Transformer, FastText, mBERT, and XLM-R baselines. For response generation, the proposed framework achieved a bilingual evaluation understudy (BLEU) score of 44.1%, a longest-common-subsequence recall-oriented understudy for gisting evaluation (ROUGE-L) score of 60.3%, and a context-retention accuracy (CRA) of 81.5%. For intent classification, it attained an F1-score of 87.3% and an area under the receiver operating characteristic curve (ROC-AUC) of 0.91; for context-dependency detection, it achieved an F1-score of 84.3%. The framework also reduced inference latency and was robust to code-switching and noisy conversational input. Human evaluation confirmed improvements in response clarity, cultural relevance, and multi-turn coherence. The findings show that hybrid embeddings combined with multilevel attention provide an effective and scalable approach to conversational AI for LRLs, with potential applicability to other African languages.
G. C. Uzoaru, I. Ayogu, J. N. Odii et al.· Journal of the Nigerian Soci...· 0 citations
Abstractive text summarization is a critical yet challenging task in natural language processing, requiring models to generate concise, coherent, and factually accurate summaries. Existing transformer-based architectures, such as BART and T5, often underperform in scenarios where language semantics evolve over time or when lexical diversity is limited. These limitations result in summaries that lack contextual relevance and semantic generalization, especially in dynamic domains such as news reporting. In this work, we present a lightweight hybrid abstractive summarization model that enhances temporal awareness and linguistic flexibility. The model integrates synonym-based augmentation and time-sensitive embeddings into a base transformer encoder-decoder architecture, enabling the generation of temporally aligned and semantically enriched summaries. To further improve reliability, we introduce a factual alignment evaluation module that assesses entity-level consistency using named entity recognition. We conduct extensive experiments on two benchmark datasets, XSum and CNN/Daily Mail, and evaluate the model using a wide range of metrics, including ROUGE, BLEU, METEOR, BERTScore, and factual consistency checks. The ablation study confirms the individual contributions of temporal modeling and synonym enrichment, with the combined model achieving up to 8.3% ROUGE-1 and 7.9% METEOR improvements over the baseline. The model demonstrates practical applicability for summarization systems requiring both temporal relevance and semantic robustness, while maintaining CPU efficiency and interpretability. Implications for downstream tasks like news trend analysis are also explored.
K. Abdalgader, Falah Y. H. Ahmed, Blen Wassie et al.· Discover Artificial Intellig...· 0 citations
Several new trends related to the rising volume of multilingual digital content have spurred a greater need for Natural Language Processing (NLP) systems capable of capturing the latent themes and understanding sentiment without stripping out the structure within a specific language. In this study, we introduce an integrated framework based on transformer architecture for topic modeling and sentiment analysis for Hindi and Italian social-media discourse, customer reviews and news corpus. The framework integrates language-aware representations in context, the sentiment classification system of supervised learning and the BERTopic-based topic discovery system. Hindi Processing focuses on Devanagari normalization, transliteration, mixed code and Hindi-English while Italian Processing focuses on inflectional, clitic morphology, colloquial and regionalization variation. In addition to classical baselines (Support Vector Machines (SVM), Long Short-Term Memory (LSTM) networks, and Latent Dirichlet Allocation (LDA)), MuRIL, UmBERTo, AlBERTo and LLaMAntino are studied. The performance is assessed descriptively in terms of accuracy, F1 Score, topic coherence (Cᵥ), and training behavior and out of distribution error analysis. On sentiment classification using Hindi, MuRIL achieved an accuracy of 0.89 and an F1-score of 0.88 which were better than the best reported sentiment classification in classical Hindi by 0.08 on both measures. UmBERTo's results for Italian were 0.90 accuracy and 0.89 F1-score, while LLaMAntino obtained the best reported sentiment scores of 0.91 and 0.90. For Hindi, the relative improvement over LDA was 53.7% and for Italian, it was 50.0% with respect to the topic coherence score measured by BERTopic. The findings suggest that there is clear benefit for morphologically complex text and mixed script text for using contextual embeddings and language-specific pretraining. However, there are still some limitations such as dialect-sensitive mistakes, spurious correlations, computational cost and low interpretability. Finally, given that error and fairness audits are not widely available, the study offers a more comprehensive grounding in error and fairness for the responsible deployment of multilingual topic and sentiment systems.
Sunita Basalingayya, T. J. Peter· Journal of Intelligent Decis...· 0 citations
Integrating structured knowledge graphs (KGs) with Large Language Models (LLMs) is essential for trustworthy, knowledge intensive conversational systems. However, existing Retrieval Augmented Generation (RAG) methods typically rely on a retrieval-as-context paradigm that linearizes structured subgraphs into unstructured prompt tokens. This approach not only flattens rich structural dependencies but also leads to context inflation and evidence attenuation in multi-turn dialogues. To address these limitations, we propose KGA-LM, a framework that integrates external knowledge via representation-level grounding. Rather than treating retrieved evidence as transient input artifacts, KGA-LM encodes compact multi-hop subgraphs using a Graph Transformer and fuses them into the LLM decoder through a compatibility-aware latent interface. This design aligns the heterogeneous latent spaces of the graph encoder and the LLM, while a dual-gated fusion mechanism dynamically regulates the influence of non-parametric graph evidence across turns. Experiments on multiple conversational benchmarks demonstrate that KGA-LM significantly improves factual accuracy and reduces hallucination compared to prompt-linearized baselines. Crucially, by decoupling knowledge injection from prompt length, our approach mitigates retrieval signal decay under long contexts, offering a superior trade-off between grounding quality and inference efficiency.
Yunfei Li, Chengfei Liu, Rui Zhou et al.· Proceedings of the 32nd ACM...· 0 citations
Large language models (LLMs) have sparked renewed interest in Conversational Information Retrieval (CIR). Within this shift, Conversational Dataset Retrieval (CDR) is emerging as a new subfield that focuses on using natural, context-aware dialogue to discover structured and semi-structured datasets. We present the first integrative review of this rapidly evolving landscape. We synthesise insights from 44 publications spanning user studies, conversational system design choices, dataset representation and access, and evaluation methods. We introduce a novel, layered conceptual framework that organizes research in CDR across four key dimensions: User Layer, System Layer, Data Layer, and Evaluation Layer. This framework is used to highlight recurring design patterns, technical advances, and persistent gaps. Our analysis identifies core challenges in CDR, including the lack of standardized evaluation benchmarks and limited support for ambiguous or evolving user intent. The aim of this survey is to provide a structured foundation for future research on CDR and guide the development of more interactive, intelligent, and user-centric dataset retrieval systems.
Lisa-Yao Gan, Johanna Walker, E. Simperl et al.· Information Systems Frontier...· 0 citations
The exponential growth of social media has reshaped global communication and decision‐making in business, politics and economics. Yet, the sheer volume and informal, unstructured nature of user‐generated content present major challenges for meaningful analysis. This study introduces a novel abstractive summarisation framework designed to distill coherent and semantically rich summaries from social media discussions. Built on the T5 transformer architecture and enhanced through targeted transfer learning, the system effectively captures the fragmented, slang‐rich language patterns common across platforms. Evaluation is conducted using a suite of semantic‐aware metrics—including ROUGE‐WE, SUPERT and Shannon entropy—alongside human‐centric criteria such as coherence, fluency, consistency and lexical diversity. Results show that the proposed model consistently outperforms mainstream summarisation methods, including advanced systems like ChatGPT, particularly in preserving semantic alignment and improving readability under noisy conditions. Comparative analysis underscores the framework's robustness in handling unstructured, domain‐specific discourse. These findings position the model as a valuable tool for real‐time, high‐volume social media analytics. Future work will explore hybrid neural quality assessment and interactive feedback mechanisms to further enhance domain adaptability and summary fluency.
A. Papagiannopoulou, C. Angeli· Expert systems· 0 citations