Large language models face significant cost challenges in long-sequence inference. To address this, reusing historical Key-Value (KV) Cache for improved inference efficiency has become a mainstream approach. Recent advances further enhance throughput by sparse attention mechanisms to select the most relevant KV Cache, thereby reducing sequence length. However, such techniques are limited to single-context scenarios, where historical KV Cache is computed sequentially with causal-attention dependencies. In retrieval-augmented generation (RAG) scenarios, where retrieved documents as context are unknown beforehand, each document’s KV Cache is computed and stored independently (termed multiple-context KV Cache), lacking cross-attention between contexts. This renders existing methods ineffective. Although prior work partially recomputes multiple-context KV Cache to mitigate accuracy loss from missing cross-attention, it requires retaining all KV Cache throughout, failing to reduce memory overhead. This paper presents SamKV, the first exploration of attention sparsification for multiple-context KV Cache. Specifically, SamKV takes into account the complementary information of other contexts when sparsifying one context, and then locally recomputes the sparsified information. Experiments demonstrate that our method compresses sequence length to 15% without accuracy degradation compared with full-recomputation baselines, significantly boosting throughput in multi-context RAG scenarios.

Compositional reasoning is a critical capability for multimodal models, enabling systematic understanding of complex scenes through structured combinations of objects, attributes, and relations. However, existing research on this ability primarily focuses on vision-language models (VLMs, e.g., CLIP and SigLIP), with limited exploration of multimodal large language models (MLLMs). To address this gap, we introduce CR³, a novel framework that enhances compositional reasoning abilities of MLLMs via rule-based reinforcement learning. CR³ leverages rule-based rewards to optimize the MLLM's policy on systematically curated multimodal instruction-following tasks, guided by a model-adaptive dynamic task mixing strategy. Our approach boosts performance by over 19% on three compositional reasoning benchmarks, significantly outperforming supervised fine-tuning (SFT) by at least 12%. Crucially, CR³ demonstrates superior generalization by improving performance on out-of-domain benchmarks where SFT methods degrade, highlighting its effectiveness and data efficiency.

Yumeng Fu, Weitao Huang, Junjie Wu, Hao Teng, Shouduo Shang, Meishan Zhang, Bingquan Liu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Emotion Recognition in Conversation (ERC), the task of identifying the emotion of each utterance in a conversation, is crucial for humanmachine interaction.Existing LLM-based ERC methods focus on standard prompting and slow thinking for emotion analysis.However, they suffer from the lack of human-like emotion reasoning and discrimination between similar emotions, thus limiting accurate emotion predictions.To this end, we present JoPR, jointing perception-curriculum learning and emotional reasoning for conversational emotion recognition.Specifically, we devise a multi-dimension curriculum with long CoT fine-tuning to clone human-like emotion reasoning.We further design an emotion-specific reward function in a novel reinforcement learning framework, thereby enhancing the discernment between similar emotions.We conduct extensive experiments on three widely used benchmark datasets, and the results demonstrate that our JoPR achieves consistent and significant improvements over baselines.

Distinguishing spatial relations is a basic part of human cognition which requires fine-grained perception on cross-instance. Although benchmarks like MME, MMBench and SEED comprehensively have evaluated various capabilities which already include visual spatial reasoning(VSR). There is still a lack of sufficient quantity and quality evaluation and optimization datasets for Vision Large Language Models(VLLMs) specifically targeting visual positional reasoning. To handle this, we first diagnosed current VLLMs with the VSR dataset and proposed a unified test set. We found current VLLMs to exhibit a contradiction of over-sensitivity to language instructions and under-sensitivity to visual positional information. By expanding the original benchmark from two aspects of tunning data and model structure, we mitigated this phenomenon. To our knowledge, we expanded spatially positioned image data controllably using diffusion models for the first time and integrated original visual encoding(CLIP) with other 3 powerful visual encoders(SigLIP, SAM and DINO). After conducting combination experiments on scaling data and models, we obtained a VLLM VSR Expert(VSRE) that not only generalizes better to different instructions but also accurately distinguishes differences in visual positional information. VSRE achieved over a 27% increase in accuracy on the VSR test set. It becomes a performant VLLM on the position reasoning of both the VSR dataset and relevant subsets of other evaluation benchmarks. We hope it will accelerate advancements in VLLM on VSR learning.

Emotion recognition in conversation (ERC), the task of discerning human emotions for each utterance within a conversation, has garnered significant attention in human-computer interaction systems. Previous ERC studies focus on speaker-specific information that predominantly stems from relationships among utterances, which lacks sufficient information around conversations. Recent research in ERC has sought to exploit pre-trained large language models (LLMs) with speaker modelling to comprehend emotional states. Although these methods have achieved the encouraging results, the extracted speaker-specific information struggles to indicate emotional dynamics. In this paper, motivated by the fact that speaker characteristics play a crucial role and LLMs have rich world knowledge, we present LaERC-S, a novel framework that stimulates LLMs to explore speaker characteristics involving the mental state and behavior of interlocutors, for accurate emotion predictions. To endow LLMs with these knowledge information, we adopt the two-stage learning to make the models reason speaker characteristics and track the emotion of the speaker in complex conversation scenarios. Extensive experiments on three benchmark datasets demonstrate the superiority of LaERC-S, reaching the new state-of-the-art.

Multimodal Emotion Recognition in Conversations (MERC) identifies utterance emotions by integrating both contextual and multimodal information from dialogue videos. Existing methods struggle to capture emotion shifts due to label replication and fail to preserve positive independent modality contributions during fusion. To address these issues, we propose a Dual Contrastive Learning Framework (DCLF) that enhances current MERC models without additional data. Specifically, to mitigate label replication effects, we construct context-aware contrastive pairs. Additionally, we assign pseudo-labels to distinguish modality-specific contributions. DCLF works alongside basic models to introduce semantic constraints at the utterance, context, and modality levels. Our experiments on two MERC benchmark datasets demonstrate performance gains of 4.67%-4.98% on IEMOCAP and 5.52%-5.89% on MELD, outperforming state-of-the-art approaches. Perturbation tests further validate DCLF’s ability to reduce label dependence. Additionally, DCLF incorporates emotion-sensitive independent modality features and multimodal fusion representations into final decisions, unlocking the potential contributions of individual modalities.

Existing cold-start recommendation methods typically use item-level alignment strategies to align the content feature and collaborative feature of warm items during model training. However, these methods are less effective for cold items with low semantic similarity to the warm items when they first appear in the test stage, as they have no historical interactions to obtain the collaborative feature. In this paper, we propose a preference aware recommendation (PARec) model with hierarchical item alignment to solve the item cold-start issue. Our approach exploits user preference from historical records to achieve group-level alignment with item content feature, enhancing recommendation performance. Specifically, our hierarchical item alignment strategy improves recommendations for both high and low similarity cold items by using item-level alignment for high similarity cold items and introducing group-level alignment for low similarity cold items. Low similarity cold items can be successfully recommended through relationships among items, captured by our group-level alignment, based on their co-occurrence possibilities and semantic similarities. For model training, a hierarchical contrastive objective function is presented to balance the performance of warm and cold items, achieving better overall performance. Extensive experiments demonstrate the effectiveness of our method, with results showing its superiority compared to state-of-the-art approaches.

Existing cold-start recommendation methods often adopt item-level alignment strategies to align the content feature and the collaborative feature of warm items for model training, however, cold items in the test stage have no historical interactions with users to obtain the collaborative feature. These existing models ignore the aforementioned condition of cold items in the training stage, resulting in the performance limitation. In this paper, we propose a preference aware dual contrastive learning based recommendation model (PAD-CLRec), where the user preference is explored to take into account the condition of cold items for feature alignment. Here, the user preference is obtained by aggregating a group of collaborative feature of the warm items in the user's purchase records. Then, a group-level alignment between the user preference and the item's content feature can be realized via a proposed preference aware contrastive function for enhancing cold-item recommendation. In addition, a joint objective function is introduced to achieve a better trade-off between the recommendation performance of warm items and cold items from both item-level and group-level perspectives, yielding better overall recommendation performance. Extensive experiments are conducted to demonstrate the effectiveness of the proposed method, and the results show the superiority of our method, as compared with the state-of-the-arts.

The Multi-Party Conversation (MPC) system has gained attention for its relevance in modern communication. Recent work has focused on developing specialized models for different MPC subtasks, improving state-of-the-art (SOTA) performance. However, since MPC demands often arise collaboratively, managing multiple specialized models is impractical. Additionally, dialogue evolves through diverse meta-information, where knowledge from specific subtasks can influence others. To address this, we propose UniMPC, a unified framework that consolidates common MPC subtasks. UniMPC uses a graph network with utterance nodes, a global node for combined local and global information, and two adaptable free nodes. It also incorporates discourse parsing to enhance model updates. We introduce MPCEval, a new benchmark for evaluating MPC systems. Experiments show UniMPC achieves over 95% of SOTA performance across all subtasks, with some surpassing existing SOTA, highlighting the effectiveness of the global node, free nodes, and dynamic discourse-aware graphs.

The fusion of language models (LMs) and knowledge graphs (KGs) is widely used in commonsense question answering, but generating faithful explanations remains challenging.Current methods often overlook path decoding faithfulness, leading to divergence between graph encoder outputs and model predictions.We identify confounding effects and LM-KG misalignment as key factors causing spurious explanations.To address this, we introduce the LM-KG Fidelity metric to assess KG representation reliability and propose the LM-KG Distribution-aware Alignment (LKDA) algorithm to improve explanation faithfulness.Without ground truth, we evaluate KG explanations using the proposed Fidelity-Sparsity Trade-off Curve.Experiments on Common-senseQA and OpenBookQA show that LKDA significantly enhances explanation fidelity and model performance, highlighting the need to address distributional misalignment for reliable commonsense reasoning.

To solve the cold-start issue that cold items have no historical interactions to obtain collaborative feature as their representation, existing methods often represent them totally based on content feature obtained from inherent content (i.e., image, video and attributes). However, these methods will lose efficacy in representing the cold item whose inherent content is much different from that of warm items, when the cold item first appears in the test stage. In this paper, we propose a mutual information assisted graph convolution network (MIGCN), which represents the cold item by simultaneously considering its inherent content and its related users’ feature captured by pair-wise mutual information (PMI). In addition, for improving the overall performance, a joint objective function that contains a relation loss and a similarity error loss is employed to achieve a better trade-off of the representation similarity between warm and cold items. Experiments conducted on the Amazon dataset demonstrate the superiority of our method, as compared with the state-of-the-art methods.

This paper proposes a new multi-modal question-answering task, named as Cross-Modal Information Complementation based Question Answering (CroMIC-QA), to promote the exploration on bridging the semantic gap between visual and linguistic signals. The proposed task is inspired by the common phenomenon that, in most user-generated QA scenarios, the information of the given textual question is incomplete, and thus it is required to merge the semantics of both the text and the accompanying image to infer the complete real question. In this work, the CroMIC-QA task is first formally defined and compared with the classic Visual Question Answering (VQA) task. On this basis, a specified dataset, CroMIC-QA-Agri, is collected from an online QA community in the agriculture domain for the proposed task. A group of experiments is conducted on this dataset, with the typical multi-modal deep architectures implemented and compared. The experimental results show that the appropriate text/image presentations and text-image semantic interaction methods are effective to improve the performance of the framework.

Multi-hop questions are associated with a series of justifications, and one needs to obtain the answers by following the reasoning path (RP) that orders the justifications adequately. So reasoning path retrieval becomes a critical preliminary stage for multi-hop Question Answering (QA). Within the RP, two fundamental challenges emerge for better performance: (i) what the order of the justifications in the RP should be, and (ii) what if the wrong justification has been in the path. In this paper, we propose Reasoning Path Augmentation (RPA), which uses reasoning path reordering and augmentation to handle the above two challenges, respectively. Reasoning path reordering restructures the reasoning by targeting the easier justification first but difficult one later, in which the difficulty is determined by the overlap between query and justifications since the higher overlap means more lexical relevance and easier searchable. Reasoning path augmentation automatically generates artificial RPs, in which the distracted justifications are inserted to aid the model recover from the wrong justification. We build RPA with a naive pre-trained model and evaluate RPA on the QASC and MultiRC datasets. The evaluation results demonstrate that RPA outperforms previously published reasoning path retrieval methods, showing the effectiveness of the proposed methods. Moreover, we present detailed experiments on how the orders of justifications and the percent of augmented paths affect the question- answering performance, revealing the importance of polishing RPs and the necessity of augmentation.

In recent years, neural networks have achieved impressive performance on dialogue response generation. However, most of these models still suffer from some shortcomings, such as yielding uninformative responses and lacking explainable ability. This article proposes a Two-stage Dialogue Response Generation model (TSRG), which specifies a method to generate diverse and informative responses based on an interpretable procedure between stages. TSRG involves a two-stage framework that generates a candidate response first and then instantiates it as the final response. The positional information and a resident token are injected into the candidate response to stabilize the multi-stage framework, alleviating the shortcomings in the multi-stage framework. Additionally, TSRG allows adjusting and interpreting the interaction pattern between the two generation stages, making the generation response somewhat explainable and controllable. We evaluate the proposed model on three dialogue datasets that contain millions of single-turn message-response pairs between web users. The results show that, compared with the previous multi-stage dialogue generation models, TSRG can produce more diverse and informative responses and maintain fluency and relevance.

Previous works show that Pre-trained Language Models (PLMs) can capture factual knowledge.However, some analyses reveal that PLMs fail to perform it robustly, e.g., being sensitive to the changes of prompts when extracting factual knowledge.To mitigate this issue, we propose to let PLMs learn the deterministic relationship between the remaining context and the masked content.The deterministic relationship ensures that the masked factual content can be deterministically inferable based on the existing clues in the context.That would provide more stable patterns for PLMs to capture factual knowledge than randomly masking.Two pre-training tasks are further introduced to motivate PLMs to rely on the deterministic relationship when filling masks.Specifically, we use an external Knowledge Base (KB) to identify deterministic relationships and continuously pre-train PLMs with the proposed methods.The factual knowledge probing experiments indicate that the continuously pre-trained PLMs achieve better robustness in factual knowledge capturing.Further experiments on question-answering datasets show that trying to learn a deterministic relationship with the proposed methods can also help other knowledge-intensive tasks.

Collecting supporting evidence from large corpora of text (e.g., Wikipedia) is of great challenge for open-domain Question Answering (QA). Especially, for multi-hop open-domain QA, scattered evidence pieces are required to be gathered together to support the answer extraction. In this paper, we propose a new retrieval target, hop, to collect the hidden reasoning evidence from Wikipedia for complex question answering. Specifically, the hop in this paper is defined as the combination of a hyperlink and the corresponding outbound link document. The hyperlink is encoded as the mention embedding which models the structured knowledge of how the outbound link entity is mentioned in the textual context, and the corresponding outbound link document is encoded as the document embedding representing the unstructured knowledge within it. Accordingly, we build HopRetriever which retrieves hops over Wikipedia to answer complex questions. Experiments on the HotpotQA dataset demonstrate that HopRetriever outperforms previously published evidence retrieval methods by large margins. Moreover, our approach also yields quantifiable interpretations of the evidence collection process.

Although the proper use of idioms can enhance the elegance of writing, the active use of various expressions is a challenge because remembering idioms is difficult.In this study, we address the problem of idiom recommendation by leveraging a neural machine translation framework, in which we suppose that idioms are written in one pseudo target language.Two types of reallife datasets are collected to support this study.Experimental results show that the proposed approach achieves promising performance compared with other baseline methods.

In human-to-human conversations, the context generally provides several backgrounds and strategic points for the following response. Therefore, many response generation approaches have explored the methodologies to incorporate the context into the encoder-decoder architecture, to generate context-aware responses that are remarkably relevant and cohesive to the given context. However, most approaches pay less attention to semantic interactions implicitly existing within contextual utterances, which are of great importance to capture semantic clues of the given dialog context, indeed. This paper proposes a dynamic working memory mechanism to model long-term semantic hints in the conversation context, by performing semantic interactions between utterances and updating context representation dynamically. Then, the outputs of the dynamic working memory are employed to provide helpful clues for the encoder-decoder architecture to generate responses to the given dialog. We have evaluated the proposed approach on Twitter Customer Service Corpus and OpenSubtitles Corpus, with several automatic evaluation metrics and the human evaluation, and the empirical results show the effectiveness of the proposed method.

Zhen Xu, Nan Jiang, Bingquan Liu, Wenge Rong, Bowen Wu, Baoxun Wang, Zhuoran Wang, Xiaolong Wang. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers). 2018.

Automatic chatbots (also known as chat-agents) have attracted much attention from both researching and industrial fields. Generally, the semantic relevance between users' queries and the corresponding responses is considered as the essential element for conversation modeling in both generation and ranking based chat systems. By contrast, it is a nontrivial task to adopt the users' information, such as preference, social role, etc., into conversational models reasonably, while users' profiles play a significant role in the procedure of conversations by providing the implicit contexts. This paper aims to address the personalized response ranking task by incorporating user profiles into the conversation model. In our approach, users' personalized representations are latently learned from the contents posted by them via a two-branch neural network. After that, a deep neural network architecture is further presented to learn the fusion representation of posts, responses, and personal information. In this way, the proposed model could understand conversations from the users' perspective; hence, the more appropriate responses are selected for a specified person. The experimental results on two datasets from social network services demonstrate that our approach is hopeful to represent users' personal information implicitly based on user generated contents, and it is promising to perform as an important component in chatbots to select the personalized responses for each user.

This paper presents a Generative Adversarial Network (GAN) to model singleturn short-text conversations, which trains a sequence-to-sequence (Seq2Seq) network for response generation simultaneously with a discriminative classifier that measures the differences between human-produced responses and machinegenerated ones.In addition, the proposed method introduces an approximate embedding layer to solve the non-differentiable problem caused by the sampling-based output decoding procedure in the Seq2Seq generative model.The GAN setup provides an effective way to avoid noninformative responses (a.k.a "safe responses"), which are frequently observed in traditional neural response generators.The experimental results show that the proposed approach significantly outperforms existing neural response generation models in diversity metrics, with slight increases in relevance scores as well, when evaluated on both a Mandarin corpus and an English corpus.

It is critical for automatic chat-bots to gain the ability of conversation comprehension, which is the essence to provide context-aware responses to conduct smooth dialogues with human beings. As the basis of this task, conversation modeling will notably benefit from the background knowledge, since such knowledge indeed implicates semantic hints that help to further clarify the relationships between sentences within a conversation. In this paper, a deep neural network is proposed to incorporate background knowledge for conversation modeling. Through a recall mechanism with a specially designed recall-gate, background knowledge as global memory can be motivated to cooperate with local cell memory of Long Short-Term Memory (LSTM), so as to enrich the ability of LSTM to capture the implicit semantic clues in conversations. In addition, this paper introduces the loose-structured domain knowledge as background knowledge, which can be built with slight amount of manual work and easily adopted by the recall-gate. Our model is evaluated on the context-oriented response selecting task, and experimental results on two datasets have shown that our approach is promising for modeling conversations and building key components of automatic chat systems.

Facing a large amount of entities appearing on the web, entity linking has recently become useful. It assigns an entity from a resource to one name mention to help users grasp the meaning of this name mention. Unfortunately, many possible entities can be assigned to one name mention. Apparently, the usually co-occurring name mentions are related and can be considered together to determine their best assignments. This approach is called collective entity linking and is often conducted based on entity graph. However, traditional collective entity linking methods either consume much time due to the large scale of entity graph or obtain low accuracy due to simplifying graph. To improve both accuracy and efficiency, this article proposes a novel collective entity linking algorithm. It first constructs an entity graph by connecting any two related entities, and then a probability-based objective function is proposed on this graph to ensure the high accuracy of the linking result. Via this function, we convert entity linking to the process of finding the nodes with the highest PageRank Values. Greedy search and an adjusted Monte Carlo random walk are proposed to fulfill this work. Experimental results demonstrate that our algorithm performs much better than traditional linking methods.

Drug Entity Recognition (DER) is a crucial task for information extraction in biomedical text. Much of previous work for DER using known drugs to build features, however, the known drug resources are limited. In this paper, we proposed a semi-supervised learning to extend an existing drug dictionary. With the extended dictionary, the features for DER can be enriched. Using Conditional Random Fields (CRF) model with the enriched features, an F-measure of 89.26% is achieved on DDIExtraction2013 challenge data set, which outperforms the best system of the DDIExtraction 2013 challenge.

The link prediction problem in social networks is to estimate the value of the link that can represent relationship between social members. Researchers have proposed several methods for solving link prediction and a number of features have been used. Most of these models are learned with only considering the features from one kind of data. In this paper, by considering the data from link network structure and user comment, both of which could imply the concept of link value, we propose multimodal learning based approaches to predict the link values. The experiment results done on dataset from typical social networks show that our model could learn the joint representation of these datas properly, and the method MDBN outperforms other state-of-art link prediction methods.

With the development of social media websites, more and more users start to show their attitudes and emotions to each other. Some of these interactions can be represented as links with sign values(positive or negative). In this paper, a unified method is proposed for link prediction and feature analysis. This paper focuses on the data from social media websites and tries to find the features that determine the sign value mostly. Based on the features extracted from the users' self statuses and from their relationships with neighbors, our method can predict the links' values with high accuracy. By analyzing the models generated over different datasets, our experiments find out the common determining features for link prediction. Based on our results, advices on how to predict links' values and get more positive links in future are given to users.

The human-generated question-answer pairs in the Web social communities are of great value for the research of automatic question-answering technique. Due to the large amount of noise information involved in such corpora, it is still a problem to detect the answers even though the questions are exactly located. Quantifying the semantic relevance between questions and their candidate answers is essential to answer detection in social media corpora. Since both the questions and their answers usually contain a small number of sentences, the relevance modeling methods have to overcome the problem of word feature sparsity. In this article, the deep learning principle is introduced to address the semantic relevance modeling task. Two deep belief networks with different architectures are proposed by us to model the semantic relevance for the question-answer pairs. According to the investigation of the textual similarity between the community-driven question-answering (cQA) dataset and the forum dataset, a learning strategy is adopted to promote our models’ performance on the social community corpora without hand-annotating work. The experimental results show that our method outperforms the traditional approaches on both the cQA and the forum corpora.

Although Support Vector Machines(SVM) succeed in classifying several image databases using image descriptors proposed in the literature, no single descriptor can be optimal for general object categorization. This paper describes a novel framework to learn the optimal combination of kernels corresponding to multiple image descriptors before SVM training, leading to solve a quadratic programming problem efficiently. Our framework takes into account the variation of kernel matrix and imbalanced dataset, which are common in real world image categorization tasks. Experimental results on Graz-01 and Caltech-101 image databases show the effectiveness and robustness of our algorithm.

Extracting question-answer pairs from online forums is a meaningful work due to the huge amount of valuable user generated resource contained in forums. In this paper we consider the problem of extracting Chinese question-answer pairs for the first time. We present a strategy to detect Chinese questions and their answers. We propose a sequential rule based method to find questions in a forum thread, then we adopt non-textual features based on forum structure to improve the performance of answer detecting in the same thread. Experimental results show that our techniques are very effective.

Domain-specific term extraction contributes to all domain-oriented natural language processing tasks. Given a small set of domain-specific terms as seed terms, new terms from unlabeled corpora can be extracted by bootstrapping a term classifier to discover the association between seed terms and new terms. Traditional term representation method for domain-specific term extraction represents a term in a feature space of documents, which depicts association of terms which share common documents. This representation can't depict the inner-document information of terms and requires extracted terms to occur in multiple documents. A new term representation method in global contextual space is proposed for domain-specific term extraction in this paper. This representation mechanism depicts the association of terms which share common global contexts. The information of terms within certain document and among corpora is depicted by global contexts. Experiments on Chinese web corpus show that the proposed domain-specific term extraction method with global contextual representation outperforms traditional method with representation mechanism in documents space. The improvement for low frequency terms is much higher for the proposed method.

Sequence labeling is a core task in natural language processing. The maximum entropy Markov model (MEMM) is a powerful tool in performing this task. This article enhances the traditional MEMM by exploiting the positional information of language elements. The stationary hypothesis is relaxed in MEMM, and the nonstationary MEMM (NS-MEMM) is proposed. Several related issues are discussed in detail, including the representation of positional information, NS-MEMM implementation, smoothing techniques, and the space complexity issue. Furthermore, the asymmetric NS-MEMM presents a more flexible way to exploit positional information. In the experiments, NS-MEMM is evaluated on both the Chinese and the English pos-tagging tasks. According to the experimental results, NS-MEMM yields effective improvements over MEMM by exploiting positional information. The smoothing techniques in this article effectively solve the NS-MEMM data-sparseness problem; the asymmetric NS-MEMM is also an improvement by exploiting positional information in a more flexible way.