The extraction of speech emotion features—particularly at the utterance level—constitutes a critical yet challenging aspect of Speech Emotion Recognition (SER). Emotion represents high-level paralinguistic information, and not all segments within a speech signal carry emotionally salient cues, especially in longer utterances. Conventional methods that process entire utterances may thus waste computational resources and introduce irrelevant acoustic interference. To address this, we propose AdaPre-Selection, an adaptive pre-selection mechanism designed to identify and extract emotion-rich segments from speech signals. Acting as a flexible front-end compression module, AdaPre-Selection consists of two complementary components: an Active Emotion Positioning (AEP) module and a Passive Emotion Constraint (PEC) module. Within AEP, Temporal Length Selection (TLS) and Start Time Point Selection (STPS) operate jointly to adaptively locate the optimal emotional segment in each utterance. Evaluated on two benchmark datasets (IEMOCAP and MSP-IMPROV) using four state-of-the-art SER models, AdaPre-Selection consistently outperforms common preprocessing baselines and delivers the most substantial improvement in recognition performance.
Log anomaly detection is critical for ensuring the reliability of large-scale distributed systems. Existing methods often fail to balance adaptability to concept drift with the interpretability required for effective root cause analysis. Deep learning models generally lack transparency, while Large Language Model (LLM)-based approaches are prone to hallucination and exhibit high latency. To overcome these challenges, this paper proposes a neuro-symbolic framework, PGCLR+SESAM, which combines two complementary modules for robust log analysis. The first module, Prompt-Guided Causal Log Reasoning (PGCLR), uses an LLM (LLaMA-3.3-70B) and prompt-based causal reconstruction to infer multi-hop dependencies among log events and generate explicit root-cause reasoning chains. The second module, Self-Evolving Semantic Anomaly Memory (SESAM), implements an active memory-guided prompting mechanism. In contrast to passive storage, SESAM dynamically retrieves and injects semantic summaries of long-tail anomalies to refine the reasoning context and enhance generalisation under distribution drift. Extensive experiments on two large-scale datasets, LogHub~2.0 (15M entries) and LogLAB (9M entries), demonstrate that this approach outperforms previous methods, including specialized LLMs such as LLM-LADE, achieving an F1-score of 0.961, maintaining real-time inference efficiency with less than 150 ms per sequence. Automated interpretability metrics, including Graph Edit Distance and Semantic Consistency Score, as well as human evaluation, are used to quantitatively assess the structural accuracy of the generated explanations. The findings demonstrate that the proposed framework enables transparent causal reasoning and offers a practical solution for trustworthy, self-adaptive system reliability management.
Speech Emotion Recognition (SER) is an important but challenging research topic in affective computing, aiming to identify the emotional state of a speaker from their speech. Emotional information in an utterance is carried by two complementary channels: the semantic content of the words (what is said) and the paralinguistic manner of delivery (how it is said). Self-supervised speech models such as WavLM capture acoustic patterns well, but they do not explicitly model lexical-semantic emotional cues. Multimodal alternatives, on the other hand, depend on automatic speech recognition (ASR) transcripts at inference time, increasing the deployment cost. To address these issues, we propose a LoRA-efficient multi-task learning framework built on a WavLM backbone, in which the semantic and paralinguistic representations of emotional speech are learned jointly inside a single speech encoder. The framework trains three objectives together: semantic distillation, paralinguistic self-reconstruction, and emotion classification, where the latter is treated as the primary task and the other two provide supporting supervision. Experiments on MSP-IMPROV and IEMOCAP validate the effectiveness of the proposed framework, and on IEMOCAP the model outperforms several existing SER methods, confirming the rationality of the design.