Key-value Memory Networks Research Articles

Enhanced Dynamic Key-Value Memory Networks for Personalized Student Modeling and Learning Ability Classification

Enhanced Dynamic Key-Value Memory Networks for Personalized Student Modeling and Learning Ability Classification

Explore Bayesian analysis in Cognitive-aware Key–Value Memory Networks for knowledge tracing in online learning

Knowledge Tracing is a crucial aspect of personalized learning that aims to track the evolving knowledge states of students with respect to one or more concepts. However, existing literature either models knowledge state for each predefined concept separately, potentially overlooking the underlying interrelationships between concepts, or lacks the capability to uncover long-term dependencies within a historical sequence of questions. Furthermore, they have limitations in precisely identifying the mastery level of individual learners for specific concepts. In order to address these problems, we propose a novel deep learning-based knowledge tracing model, named Bayesian Cognitive-aware Key–Value Memory Network (BCKVMN). BCKVMN combines the strengths of Dynamic Key–Value Memory Network (DKVMN), Long Short-Term Memory (LSTM), embedding layer, and Bayesian theory. To capture long-term dependencies in question sequences and contextual information, we extend DKVMN by incorporating a standard LSTM module that integrates cognitive processes observed in human learning. Additionally, we employ an embedding layer to uncover the influence of question difficulty on final predictions. Furthermore, we introduce Bayesian theory to enhance parameter interpretation for the model. To investigate complex representation of human learning, our approach considers not only three distinct learning states during the learning process but also the relationships between underlying concepts and students’ mastery levels for each concept. Experimental evaluations are conducted on the three real-world datasets Algebra200, ASSISTMENTS2009 and ASSISTMENTS2017. The results have demonstrated that the proposed model outperforms state-of-the-art methods (achieving nearly 7% improvement in AUC in some cases) while achieving a better balance between performance and interpretability.

Exploring the Teaching Mode of Secondary English Education Based on Big Data Technology

Abstract With the rapid development of information technology represented by big data, all walks of life are facing the inevitable trend of transformation in the direction of informationization. English education teaching, as an important part of secondary education, naturally needs to utilize big data technology to carry out in-depth reform. This paper proposes a teaching model for secondary English education that is based on big data technology. For the teaching evaluation stage of this model, we created the DKVMN-FMF teaching effect assessment model, which incorporates the forgetting mechanism. By incorporating a dynamic key-value memory network, the assessment model can predict students’ answers’ accuracy, taking into account the forgetting phenomenon in the learning process. Through comparative experiments, this model’s AUC and accuracy are higher than other models on different data sets, indicating that the DKVMN-FMF model can accurately assess students’ learning effects. During the teaching experiment, the mean English score increased to 77.014 points after implementing the big data technology-based English teaching model, marking a significant difference from the pre-experiment mean score of 65.532. The paired samples test revealed a significant difference with a Sig value of 0.001, less than 0.05. The English teaching model that utilizes big data technology can enhance students’ English performance, as demonstrated.

DKVMN-KAPS: Dynamic Key-Value Memory Networks Knowledge Tracing With Students’ Knowledge-Absorption Ability and Problem-Solving Ability

DKVMN-KAPS: Dynamic Key-Value Memory Networks Knowledge Tracing With Students’ Knowledge-Absorption Ability and Problem-Solving Ability

Confidence-based dynamic cross-modal memory network for image aesthetic assessment

Image aesthetic assessment (IAA) aims to design algorithms that can make human-like aesthetic decisions. Due to its high subjectivity and complexity, visual information alone is limited to fully predict the aesthetic quality of an image. More and more researchers try to use complementary information from user comments. However, user comments are not always available due to various technical and practical reasons. Therefore, it is necessary to find a way to reconstruct the missing textual information for aesthetic prediction with visual information only. This paper solves this problem by proposing a Confidence-based Dynamic Cross-modal Memory Network (CDCM-Net). Specifically, the proposed CDCM-Net consists of two key components: Visual and Textual Memory (VTM) network and Confidence-based Dynamical Multi-modal Fusion module (CDMF). VTM is based on the key–value memory network. It consists of a visual key memory and a textual value memory. The visual key memory learns the visual information. While the textual value memory learns to remember the textual feature and align them with the corresponding visual features. During inference, textual information can be reconstructed using only visual features. Furthermore, a CDMF module is introduced to perform trustworthy fusion. CDMF evaluates modality-level informativeness and then dynamically integrates reliable information. Extensive experiments are performed to demonstrate the superiority of the proposed method.

Use of a Deep Learning Approach for the Evaluation of Students' Online Learning Cognitive Ability

The cognitive ability evaluation of college students' online learning still faces many challenges. Evaluation of college students' online learning cognitive ability is helpful to monitor the learning effect and progress of students, so as to detect problems in time, make suggestions and take corresponding measures to improve. However, the existing evaluation methods for the cognitive ability of college students' online learning often only focus on the degree of knowledge mastery of students, while ignoring the performance of students' learning strategies, autonomous learning ability and emotional attitude in the online learning process. And it is often impossible to understand the learning status of students in real time. To this end, this article studies the evaluation method of college students’ online learning cognitive ability based on deep learning. The evaluation of online learning cognitive ability of college students is divided into two parts: online learning ability evaluation and learning recursive ability evaluation. By introducing the learning ability layer and the learning recursive ability layer into the dynamic key-value memory network and constructing the corresponding evaluation model, it’s better to evaluate the online learning cognitive ability of college students. Based on the historical learning behavior data of students in the online learning process, the online learning cognitive state is predicted, and the performance of the model is further improved through data enhancement and self-attention mechanism. Experimental results verify the validity of the constructed model.

Document-Level Event Role Filler Extraction Using Key-Value Memory Network

Previous work has demonstrated that end-to-end neural sequence models work well for document-level event role filler extraction. However, the end-to-end neural network model suffers from the problem of not being able to utilize global information, resulting in incomplete extraction of document-level event arguments. This is because the inputs to BiLSTM are all single-word vectors with no input of contextual information. This phenomenon is particularly pronounced at the document level. To address this problem, we propose key-value memory networks to enhance document-level contextual information, and the overall model is represented at two levels: the sentence-level and document-level. At the sentence-level, we use BiLSTM to obtain key sentence information. At the document-level, we use a key-value memory network to enhance document-level representations by recording information about those words in articles that are sensitive to contextual similarity. We fuse two levels of contextual information by means of a fusion formula. We perform various experimental validations on the MUC-4 dataset, and the results show that the model using key-value memory networks works better than the other models.

Kcr-FLAT: A Chinese-Named Entity Recognition Model with Enhanced Semantic Information.

The performance of Chinese-named entity recognition (NER) has improved via word enhancement or new frameworks that incorporate various types of external data. However, for Chinese NER, syntactic composition (in sentence level) and inner regularity (in character-level) have rarely been studied. Chinese characters are highly sensitive to sentential syntactic data. The same Chinese character sequence can be decomposed into different combinations of words according to how they are used and placed in the context. In addition, the same type of entities usually have the same naming rules due to the specificity of the Chinese language structure. This paper presents a Kcr-FLAT to improve the performance of Chinese NER with enhanced semantic information. Specifically, we first extract different types of syntactic data, functionalize the syntactic information by a key-value memory network (KVMN), and fuse them by attention mechanism. Then the syntactic information and lexical information are integrated by a cross-transformer. Finally, we use an inner regularity perception module to capture the internal regularity of each entity for better entity type prediction. The experimental results show that with F1 scores as the evaluation index, the proposed model obtains 96.51%, 96.81%, and 70.12% accuracy rates on MSRA, resume, and Weibo datasets, respectively.

Interpretable Knowledge Tracing: Simple and Efficient Student Modeling with Causal Relations

Intelligent Tutoring Systems have become critically important in future learning environments. Knowledge Tracing (KT) is a crucial part of that system. It is about inferring the skill mastery of students and predicting their performance to adjust the curriculum accordingly. Deep Learning based models like Deep Knowledge Tracing (DKT) and Dynamic Key-Value Memory Network (DKVMN) have shown significant predictive performance compared with traditional models like Bayesian Knowledge Tracing (BKT) and Performance Factors Analysis (PFA). However, it is difficult to extract psychologically meaningful explanations from the tens of thousands of parameters in neural networks, that would relate to cognitive theory. There are several ways to achieve high accuracy in student performance prediction but diagnostic and prognostic reasonings are more critical in learning science. In this work, we present Interpretable Knowledge Tracing (IKT), a simple model that relies on three meaningful features: individual skill mastery, ability profile (learning transfer across skills) and problem difficulty by using data mining techniques. IKT’s prediction of future student performance is made using a Tree Augmented Naive Bayes Classifier (TAN), therefore its predictions are easier to explain than deep learning based student models. IKT also shows better student performance prediction than deep learning based student models without requiring a huge amount of parameters. We conduct ablation studies on each feature to examine their contribution to student performance prediction. Thus, IKT has great potential for providing adaptive and personalized instructions with causal reasoning in real-world educational systems.

Knowledge interaction enhanced sequential modeling for interpretable learner knowledge diagnosis in intelligent tutoring systems

One of the fundamental tasks when providing personalized tutoring services to learners in online learning systems, such as intelligent tutoring systems and massive open online courses, is the learner knowledge diagnosis (LKD). LKD obtains the learner knowledge proficiency on skills by modeling their learning performance. Learners’ knowledge construction process is not static, but evolves overtime; hence, the evolution of learners’ knowledge proficiency must be dynamically traced. Moreover, considering the wide usage of online learning systems by large numbers of learners, the LKD task also needs to meet the requirements of large-scale assessment and interpretability to explain the diagnosed results. The existing models are either designed for static scenarios or find it difficult to explain the causality between learner performance and knowledge proficiency, as well as the item characteristics. To solve these issues, we propose herein a novel model, called the knowledge interaction-enhanced dynamic LKD (KIEDLKD), to develop learner performance, and hence, dynamically diagnose and trace the evolution of each learner’s knowledge proficiency during the exercising activities. We first propose a dynamic LKD framework by unifying the strength of the memory capacity of the key-value memory network to enhance the representation of the knowledge state during learner performance modeling and the interpretability of the Item Response Theory (IRT) to explain the learner performance in terms of knowledge proficiency and item characteristics (i.e., item difficulty and discrimination). In this framework, we diagnose and trace each learner’s knowledge proficiency on each knowledge concept (KC) over time and store them into an auxiliary memory using the key-value memory network. We further infer their general proficiencies and the IRT-based item characteristics using another neural network. Moreover, we propose the knowledge interaction concept among KCs and incorporate it into the LKD procedure to further exploit the long-term dependencies in the exercising sequences, thereby devising the KIEDLKD model. We also incorporate the learner-oriented cognitive item difficulty into our model, based on each learner’s exercising history, to adaptively model the item difficulty. Based on these factors, our KIEDLKD model can not only output the learners’ knowledge proficiency in a multi-granularity manner but also output the item characteristics, making it possible to interpret the learner performances in terms of their current knowledge states and item characteristics. Extensive experiments are conducted from six perspectives on five real-world datasets to test our model.The results of learner performance prediction demonstrate the superiority of our model on the LKD task. It can also automatically discover the underlying interaction between each pair of latent KCs, and the underlying concepts for each exercise. The ablation study verifies the contributions of each component in our model. Moreover, it can depict the evolution of learner knowledge proficiency in a multi-granularity manner and provide additional information for skill domain analysis, which enables the interpretability of our model.

An Extensible Heterogeneous Network Embedding Framework for Knowledge Tracing

Knowledge tracing is of great significance for providing better personalized learning guidance and has thus attracted extensive research attention in recent years. The task of knowledge tracing is to model students' learning process on the basis of historical exercise records and trace students' knowledge proficiency, thereby predicting students' performance on future exercises or recommending exercises for better proficiency. Existing methods focus on either the skill level or the exercise level, ignoring the relationships among exercises and Knowledge Components (KCs). The classical single-factor models include the Deep Knowledge Tracing (DKT) {model} and the Dynamic Key-Value Memory Network (DKVMN) model. Although a few models, such as the Bayesian Knowledge Tracing (BKT) model and the Knowledge Proficiency Tracing (KPT) model, utilize the Q-matrix to improve model performance, most of them ignore the interaction among KCs, not to mention models that do not use the Q-matrix. Inspired by the recent success of network embedding, this paper presents a heterogeneous network embedding framework for knowledge tracing called HNEKT that takes both exercises and KCs into account. To adapt to the application of knowledge tracing, this paper also proposes several meta-paths to generate meaningful node sequences for network embedding. Besides, it explores other side information as well to improve the extensibility and effectiveness of the proposed model. Extensive experiments on three real-world datasets demonstrate the effectiveness of the HNEKT model.

Coffee With a Hint of Data: Towards Using Data-Driven Approaches in Personalised Long-Term Interactions.

While earlier research in human-robot interaction pre-dominantly uses rule-based architectures for natural language interaction, these approaches are not flexible enough for long-term interactions in the real world due to the large variation in user utterances. In contrast, data-driven approaches map the user input to the agent output directly, hence, provide more flexibility with these variations without requiring any set of rules. However, data-driven approaches are generally applied to single dialogue exchanges with a user and do not build up a memory over long-term conversation with different users, whereas long-term interactions require remembering users and their preferences incrementally and continuously and recalling previous interactions with users to adapt and personalise the interactions, known as the lifelong learning problem. In addition, it is desirable to learn user preferences from a few samples of interactions (i.e., few-shot learning). These are known to be challenging problems in machine learning, while they are trivial for rule-based approaches, creating a trade-off between flexibility and robustness. Correspondingly, in this work, we present the text-based Barista Datasets generated to evaluate the potential of data-driven approaches in generic and personalised long-term human-robot interactions with simulated real-world problems, such as recognition errors, incorrect recalls and changes to the user preferences. Based on these datasets, we explore the performance and the underlying inaccuracies of the state-of-the-art data-driven dialogue models that are strong baselines in other domains of personalisation in single interactions, namely Supervised Embeddings, Sequence-to-Sequence, End-to-End Memory Network, Key-Value Memory Network, and Generative Profile Memory Network. The experiments show that while data-driven approaches are suitable for generic task-oriented dialogue and real-time interactions, no model performs sufficiently well to be deployed in personalised long-term interactions in the real world, because of their inability to learn and use new identities, and their poor performance in recalling user-related data.

Learning to Respond with Your Favorite Stickers

Stickers with vivid and engaging expressions are becoming increasingly popular in online messaging apps, and some works are dedicated to automatically select sticker response by matching the stickers image with previous utterances. However, existing methods usually focus on measuring the matching degree between the dialog context and sticker image, which ignores the user preference of using stickers. Hence, in this article, we propose to recommend an appropriate sticker to user based on multi-turn dialog context and sticker using history of user. Two main challenges are confronted in this task. One is to model the sticker preference of user based on the previous sticker selection history. Another challenge is to jointly fuse the user preference and the matching between dialog context and candidate sticker into final prediction making. To tackle these challenges, we propose a Preference Enhanced Sticker Response Selector (PESRS) model. Specifically, PESRS first employs a convolutional-based sticker image encoder and a self-attention-based multi-turn dialog encoder to obtain the representation of stickers and utterances. Next, deep interaction network is proposed to conduct deep matching between the sticker and each utterance. Then, we model the user preference by using the recently selected stickers as input and use a key-value memory network to store the preference representation. PESRS then learns the short-term and long-term dependency between all interaction results by a fusion network and dynamically fuses the user preference representation into the final sticker selection prediction. Extensive experiments conducted on a large-scale real-world dialog dataset show that our model achieves the state-of-the-art performance for all commonly used metrics. Experiments also verify the effectiveness of each component of PESRS.

Dynamic Key-Value Memory Networks With Rich Features for Knowledge Tracing.

Knowledge tracing is an important research topic in student modeling. The aim is to model a student's knowledge state by mining a large number of exercise records. The dynamic key-value memory network (DKVMN) proposed for processing knowledge tracing tasks is considered to be superior to other methods. However, through our research, we have noticed that the DKVMN model ignores both the students' behavior features collected by the intelligent tutoring system (ITS) and their learning abilities, which, together, can be used to help model a student's knowledge state. We believe that a student's learning ability always changes over time. Therefore, this article proposes a new exercise record representation method, which integrates the features of students' behavior with those of the learning ability, thereby improving the performance of knowledge tracing. Our experiments show that the proposed method can improve the prediction results of DKVMN.

A Dynamic Knowledge Diagnosis Approach Integrating Cognitive Features

The rapid development of data analytic technologies has advanced personalized learning and increased its popularity in K-12 education significantly. Specifically, one fundamental step in personalized learning is knowledge proficiency diagnosis which reveals blind spots in students’ knowledge. However, existing approaches to diagnosis either exploit data from a one-time assessment for the cognitive diagnosis (ignoring the previous historical interactions) or trace the knowledge state using recurrent neural networks to predict students’ future performance (ignoring the cognitive features). To this end, this study proposes a dynamic approach to knowledge diagnosis integrating cognitive features with a key-value memory network to store latent exercise information and capture long-term temporal features based on cognitive psychology. Specifically, given the characteristics of assessment data in China, our approach mainly aims to model sequence data with cognitive features, including forgetting and learning. Two corresponding gates are used to weaken the knowledge memory and strengthen the repeated knowledge memory over time, respectively, in the memory updating process. Finally, to evaluate our approach, we conducted extensive experiments on four real-world datasets collected from K-12 education. The results show that the approach can effectively process the time sequence in education, whose prediction results are better and more stable than other existing baseline models. We also conducted experiments for parameter sensitivity, different feature integration methods, and the effectiveness of cognitive features to ensure that the models achieved the best results. The application visualization further confirms the practicability of our approach in dealing with problems of dynamic knowledge diagnosis.

Improving biomedical named entity recognition with syntactic information

BackgroundBiomedical named entity recognition (BioNER) is an important task for understanding biomedical texts, which can be challenging due to the lack of large-scale labeled training data and domain knowledge. To address the challenge, in addition to using powerful encoders (e.g., biLSTM and BioBERT), one possible method is to leverage extra knowledge that is easy to obtain. Previous studies have shown that auto-processed syntactic information can be a useful resource to improve model performance, but their approaches are limited to directly concatenating the embeddings of syntactic information to the input word embeddings. Therefore, such syntactic information is leveraged in an inflexible way, where inaccurate one may hurt model performance.ResultsIn this paper, we propose BioKMNER, a BioNER model for biomedical texts with key-value memory networks (KVMN) to incorporate auto-processed syntactic information. We evaluate BioKMNER on six English biomedical datasets, where our method with KVMN outperforms the strong baseline method, namely, BioBERT, from the previous study on all datasets. Specifically, the F1 scores of our best performing model are 85.29% on BC2GM, 77.83% on JNLPBA, 94.22% on BC5CDR-chemical, 90.08% on NCBI-disease, 89.24% on LINNAEUS, and 76.33% on Species-800, where state-of-the-art performance is obtained on four of them (i.e., BC2GM, BC5CDR-chemical, NCBI-disease, and Species-800).ConclusionThe experimental results on six English benchmark datasets demonstrate that auto-processed syntactic information can be a useful resource for BioNER and our method with KVMN can appropriately leverage such information to improve model performance.

KM[formula omitted]: Visual reasoning via Knowledge Embedding Memory Model with Mutual Modulation

Visual reasoning is a special kind of visual question answering, which is essentially multi-step and compositional, and also requires intensive text-visual interaction. The most important and challenging problem of visual reasoning is to design an effective and robust visual reasoning model. To this end, there are two challenges to overcome. The first is that textual and visual information must be jointly considered to make accurate inferences about reasoning. The second is that existing deep learning-based works are often too specific to a particular task. To address these issues, we propose a knowledge memory embedding model with mutual modulation for visual reasoning. This approach learns not only knowledge-based embeddings derived from key–value memory network to make the full and joint of textual and visual information, but also exploits the prior knowledge to improve the performance with knowledge-based representation learning for applying other general reasoning tasks. Experimental results on four benchmarks show that the proposed approach significantly improves performance compared with other state-of-the-art methods, guarantees the robustness with our model. Most importantly, we apply our model to four reasoning tasks, and experimentally show that our model effectively supports relational reasoning and improves performance in several tasks and datasets.

Adversarially regularized medication recommendation model with multi-hop memory network

Medication recommendation is attracting enormous attention due to its promise in effectively prescribing medicines and improving the survival rate of patients. Among all challenges, drug–drug interactions (DDI) related to undesired duplication, antagonism, or alternation between drugs could lead to fatal side effects. Previous researches usually provide models with DDI knowledge to achieve DDI reduction. However, the mixed use of patients with different DDI rates places stringent requirements on the generalization performance of models. In pursuit of a more effective method, we propose the adversarially regularized model for medication recommendation (ARMR). Specifically, ARMR firstly models temporal information from medical records to obtain patient representations and builds a key-value memory network based on information from historical admissions. Then, ARMR carries out multi-hop reading on the memory network to recommend medications. Meanwhile, ARMR uses a GAN model to adversarially regulate the distribution of patient representations by matching the distribution to a desired Gaussian distribution to achieve DDI reduction. Comparative evaluations between ARMR and baselines show that ARMR outperforms all baselines in terms of medication recommendation, achieving DDI reduction regardless of numbers of DDI types being considered.

Knowledge-Enhanced Graph Neural Networks for Sequential Recommendation

With the rapid increase in the popularity of big data and internet technology, sequential recommendation has become an important method to help people find items they are potentially interested in. Traditional recommendation methods use only recurrent neural networks (RNNs) to process sequential data. Although effective, the results may be unable to capture both the semantic-based preference and the complex transitions between items adequately. In this paper, we model separated session sequences into session graphs and capture complex transitions using graph neural networks (GNNs). We further link items in interaction sequences with existing external knowledge base (KB) entities and integrate the GNN-based recommender with key-value memory networks (KV-MNs) to incorporate KB knowledge. Specifically, we set a key matrix to many relation embeddings that learned from KB, corresponding to many entity attributes, and set up a set of value matrices storing the semantic-based preferences of different users for the corresponding attribute. By using a hybrid of a GNN and KV-MN, each session is represented as the combination of the current interest (i.e., sequential preference) and the global preference (i.e., semantic-based preference) of that session. Extensive experiments on three public real-world datasets show that our method performs better than baseline algorithms consistently.

Hierarchical Contextualized Representation for Named Entity Recognition

Named entity recognition (NER) models are typically based on the architecture of Bi-directional LSTM (BiLSTM). The constraints of sequential nature and the modeling of single input prevent the full utilization of global information from larger scope, not only in the entire sentence, but also in the entire document (dataset). In this paper, we address these two deficiencies and propose a model augmented with hierarchical contextualized representation: sentence-level representation and document-level representation. In sentence-level, we take different contributions of words in a single sentence into consideration to enhance the sentence representation learned from an independent BiLSTM via label embedding attention mechanism. In document-level, the key-value memory network is adopted to record the document-aware information for each unique word which is sensitive to similarity of context information. Our two-level hierarchical contextualized representations are fused with each input token embedding and corresponding hidden state of BiLSTM, respectively. The experimental results on three benchmark NER datasets (CoNLL-2003 and Ontonotes 5.0 English datasets, CoNLL-2002 Spanish dataset) show that we establish new state-of-the-art results.