Information Processing & Management最新文献

Depression detection based on text analysis has emerged as a research hotspot. Existing research indicates that patients’ personalized characteristics are the primary factor contributing to differences in reported experiences, which poses challenges for automated depression detection methods. To address this, we pioneered defining the fundamental components of personalized information within the text-based depression detection field and proposed the Personalized Information Embedding (PIE) model. The model narrows the gap between generic clinical symptoms and personalized patient experiences in detection, introducing a novel method for computing personalized information representations. Then, we constructed a unique depression intervention dataset containing 108 cases of subjects, the first longitudinally gathering experimental dataset in text-based depression detection. Extensive experimental evidence demonstrates that compared to advanced models, PIE demonstrates statistically significant improvements in performance (with the maximum reductions in RMSE of 0.309 and MAE of 0.232) and generalizability (with standard deviation reductions in RMSE by 75.43% and MAE by 69.77%), and the out-of-domain generalizability of personalized information representations has been validated on two larger external datasets. Additionally, we conducted case studies to analyze how personalized information leads to improved model capabilities. This research serves as a pilot and reference for developing personalized models in text-based depression detection.

As complex multi-omics data in the medical field tend to be multi-modal. Integrating these multimodal information into novel disease diagnosis models has become challenging. However, previous methods mainly focus on single omics, which cannot effectively capture the contributions between different combinations of multi-omics information. To solve this problem, based on Raman spectroscopy, FTIR spectroscopy, and metabolomics data, this paper proposes a new Cross-modal Cross-fusion network based on the Transformer self-attention mechanism (CMACF). The research focuses on effectively combining the feature patterns of different omics for disease prediction. Specifically, by constructing the Raman-IR, Raman-metabolomic, and IR spectral-metabolomic feature pairs and reasonably focusing on the information of different combination pairs through multiple stages of feature sub-network, attention cross-fusion, bimodal interaction, and sequence interaction feature level fusion, it is interesting to find that the information contribution between different pairs is different. We conducted extensive experiments on the systemic lupus erythematosus multi-omics dataset, and the accuracy and AUC values are as high as 99.44 % and 99.98 %, respectively, with the best classification effect. The results show that CMACF can efficiently fuse multi-omics medical data, provide an efficient baseline for processing medical multimodal data, and analyze the contribution of multi-omics data fusion.

The proliferation of rumors on social networks poses a serious threat to cybersecurity, justice and public trust, increasing the urgent need for rumor detection. Existing detection methods typically treat all rumors as a single homogeneous category, neglecting the diverse semantic hierarchies within rumors. Rumors pervade various domains, each with its distinct characteristics. These methods tend to lag in expressiveness when confronted with real-world scenarios involving multiple semantic levels. Furthermore, the diversity of rumors also complicates the collection of datasets, and inevitably introduces noisy data, which hinders the correctness of the learned representations. To address these challenges, we propose a rumor detection framework with Hierarchical Prototype Contrastive Learning (HPCL). In this framework, we construct a set of dynamically updated hierarchical prototypes through contrastive learning to encourage capturing the hierarchical semantic structure within rumors. Additionally, we design a difficulty metric function based on the distance between instances and prototypes, and introduce curriculum learning to mitigate the adverse effects of noisy data. Experiments on four public datasets demonstrate that our approach achieves state-of-the-art performance. Our code is publicly released at https://github.com/Coder-HenryZa/HPCL.

Empathetic response generation endeavours to perceive the interlocutor’s emotional and cognitive states in the dialogue and express proper responses. Previous studies detect the interlocutor’s states by understanding the immediate context of the dialogue. However, these methods are at an elementary/intermediate level of empathetic understanding due to the neglect of the broader context (i.e., the situation) and its associations with the dialogue, leading to inaccurate comprehension of the interlocutor’s states. In this paper, we utilize the EMPATHETIC-DIALOGUES dataset consisting of 25k dialogues, and on this basis, we propose a Situation-Dialogue Association Model (SDAM). SDAM focuses on the broader context, i.e., the situation, and enhances the understanding of empathy from explicit and implicit associations. Regarding explicit associations, we propose a bidirectional filtering encoder. It selects relevant keywords between the situation and dialogue, learning their direct lexical relevance. For implicit associations, we use a knowledge-based hypergraph network grounded to learn convoluted connections between the situation and the dialogue. Moreover, we also introduce a simple fine-tuning approach that combines SDAM with large language models to further strengthen the empathetic understanding capability. Compared to the baseline, SDAM demonstrates superior empathetic ability. In terms of emotion accuracy, fluency, and response diversity (Distinct-1/Distinct-2), SDAM achieves improvements of 12.25 (a 30.47% increase), 0.3 (a 0.85% increase), and 0.86/1.23 (116.22% and 30.67% increases), respectively. Additionally, our variant model based on large language models exhibits better emotion recognition capability without compromising response quality, specifically achieving an improvement of 0.23 (a 0.37% increase) in emotion accuracy.

Large language models (LLMs) that exhibit a remarkable ability by in-context learning (ICL) with bilingual demonstrations have been recognized as a potential solution for machine translation. However, the process of selecting these demonstrations from vast datastores is notoriously time-consuming and inefficient. Moreover, the strategies for designing effective in-context demonstrations are not well-established. To address these critical gaps, we introduce a novel Fast and Effective approach for Demonstration Selection in-Context learning (FEDS-ICL) tailored to LLMs. Our method is designed to mainly enhance the efficiency and accuracy of translation of LLMs. Our approach revolutionizes demonstration selection by designing new product quantization technique that rapidly extracts neighboring target tokens from a strategically curated subset of sentences. This method significantly deviates from the conventional exhaustive search across entire datastores, leading to a remarkable increase in speed. Furthermore, FEDS-ICL pioneers an innovative template design for in-context demonstrations, specifically crafted to amplify the translation capabilities of multilingual LLMs. In experiments, we compare our FEDS-ICL with various existing methods on across diverse language pairs on ten different LLMs. The results reveal an up to 2.1-fold increase in selection speed and an impressive enhancement in translation accuracy, outperforming existing baselines by up to 2.0 BLEU points at least on ten different LLMs. The ablation study show the proposed product quantization and multi-view demonstration can effectively enhance the efficiency and accuracy of LLMs in machine translation. The analysis on robustness of FEDS-ICL shows that the incorporation of a greater number of demonstrations can lead a positive correlation between the quantity of contextually rich demonstrations and the translation quality of LLMs. These advancements position FEDS-ICL as a transformative methodology in the domain of machine translation and pattern analysis, marking a significant leap towards more efficient and precise machine translation.

Stance detection, a fundamental task in natural language processing, identifies user stances in texts towards specific targets. The diverse targets and ever-changing expressions make it challenging to attain comprehensive knowledge from limited data. Existing methods focus on incorporating supplementary knowledge, neglecting fusion consistency during training, which is critical for preserving the rationality of the inference. In this paper, we introduce TAP, a novel approach for few-shot stance detection. TAP extends the verbalizer hierarchically, a mapping function in prompt-tuning. Constructed using a log-odds ratio of topics and targets, the verbalizer refines candidates with prior knowledge, forming the foundation for subsequent hierarchical distillation. The hierarchical distillation, a technique based on pilot experiments on the hierarchical verbalizer, ensures the fusion of diverse knowledge during prompt-tuning, maintaining consistency throughout the training process. Notably, TAP constructs verbalizers without external knowledge augmentation. The hierarchical distillation involves a joint loss function, contributing to the model’s robustness and training consistency. Extensive experiments are conducted on SemEval2016t6 and ArgMin datasets with 13 different targets. The proposed method is evaluated on various few-shot and full-data settings with F1-Macro and F1-Micro scores. On average, TAP achieves overall improvements of 4.71% and 3.76% over state-of-the-art baselines on Semeval2016t6 and ArgMin datasets, respectively, in few-shot scenarios.

In the finance domain, nested named entities recognition has become a hot topic in named entity recognition tasks. Traditional nested entity recognition methods easily ignore the dependency relationships between entities, and these methods are mostly suitable for English general domain. Therefore, we propose a Chinese nested entity recognition method for the finance domain based on heterogeneous graph network(HGFNER). This method consists of two parts: the boundary division model of candidate entities and the internal relationship graph model of candidate entities. First, the boundary division model of candidate entities that introduces expert knowledge is used to partition the flat entities contained in the text and segment the text to address issues such as long entity boundaries and strong domain features in the Chinese finance domain. Then, by using heterogeneous graphs to represent the internal structure of entities from both spatial and syntactic dependencies to achieve the goal of learning dependency relationships between entities from multiple perspectives. Meanwhile, so as not to affect the operational efficiency of the model, we also propose a fast matching algorithm DAAC_BM for n-gram sequences in domain dictionaries to solve the problems of memory overflow and space waste faced by multi-pattern fast matching algorithms in Chinese matching. In addition, we propose a Chinese nested entity dataset CFNE for the financial field, which, as far as we know, is the first publicly available annotated dataset in the field. HGFNER achieves state-of-the-art macro-F1 value on CFNE, reaching 86.41%.

This paper examines the role of response in crowdfunding to guide fundraisers to monetize their responses better. In all, 6,405 commenters on a large crowdfunding platform in China (Modian.com) are observed. Grounded on the interaction texts, we identify, summarize, and extract the comments and the strategies for fundraisers to respond through several rounds of coding. It is found that potential investors comment to collect information or to express themselves, while fundraisers are either project-oriented or investor-oriented when responding to online reviews. We adopt the Naïve Bayes classifier to classify the crowdfunding demand of commenters, and entrepreneur's response strategy, as being either investor-oriented (60.8 %) or project-oriented (39.2 %). Econometric models are constructed to empirically test and quantify the impact of fundraiser response strategies on commenter investment decisions. The findings inform that commenters who are responded to by fundraisers are about 4 times more likely to invest than those who receive no response or receive responses from peer commenters. Also, a fundraiser project-oriented response strategy achieves 2.25 times better crowdfunding performance over an investor-oriented one. Robustness testing is performed using other criteria for classifying response strategies, as well as testing with a sample from another platform.

Self-explainable Graph Neural Networks (GNNs) provide explanations alongside their predictions, making the model transparent and facilitating their wide adoption in high-stakes tasks. Current studies on constructing such GNNs are limited by the single receptive field, resulting in the modeling of spurious correlations in self-explainable GNNs. To address this issue, this paper introduces a GNN model with incorporated multi-granularity receptive fields, capturing causal correlations during the model construction and providing explanations alongside its predictions. Specifically, we employ closeness matrices with multiple structural orders to construct multi-granularity receptive fields for the model. Subsequently, we design a model architecture with sliced channels to integrate representations learned from multiple receptive fields heuristically. Objective functions from a causal perspective are further designed to guide the optimization of the proposed model. Experiments conducted on five real-world datasets and one synthetic dataset demonstrate the superior performance of the proposed model. In terms of classification accuracy, compared to SOTA baseline, the proposed GNN achieves the improvement of 0.17%, 1.99%, 0.70%, 0.83%, and 0.78% on the real-world datasets MUTAG, PTC, PROTEINS, IMDB-M, and IMDB-B, respectively. Compared to three self-explainable baselines, qualitative and quantitative studies are conducted on MUTAG, PTC, PROTEINS, IMDB-M, IMDB-B, and the synthetic dataset Spurious-Motif. Experimental results confirm that the proposed model can accurately identify the essential substructures, such as $N{O}_2$ in the MUTAG dataset. Additionally, the proposed model assigns significant weights to the motif part and distinguishes it from the base part in the Spurious-Motif dataset, enhancing the accuracy of graph classification and the explanations of the predictions. The classifications along with explanations obtained with this approach align with human cognition and experience.

The objective of the Aspect Sentiment Triplet Extraction (ASTE) task is to identify triplets of (aspect, opinion, sentiment) from user-generated reviews. The current study does not extensively integrate the interaction between word pairs and aspect-opinion pairs during the learning process at the granularity of sentence analysis. Furthermore, the bidirectional inference for the triplet, along with the parallel computing approach for long-span texts, also fail to achieve efficient unification. We introduce a new perspective: Span-level Bidirectional Retention Scheme(SBRS) for Aspect Sentiment Triplet Extraction model. The model comprises two pathways. The first pathway involves extracting effective aspect-opinion pair outcomes via two progressive submodules that operate on words and word pairs at varying scales. Building on the first pathway, the second pathway senses the interaction information of word pairs through bidirectional recursion and combines an efficient parallel computing approach. This combination allows the model to utilize three features – context, semantics, and relationship – to accurately identify the sentimental orientation. Thus, the two pathways facilitate the learning of relation-aware representations of word pairs. We carried out experiments on two public datasets, showing an average enhancement of 3.34% and 1.72% in F1 scores compared to the most recent baselines models, and multiple experiments from diverse angles proved the model’s superiority.