The intersection of psychology and large language models (LLMs) has attracted increasing attention in recent research. Prior studies have shown that incorporating psychological prompts into task instructions can influence LLM performance, typically from a user-centered perspective as external incentives. In contrast, internal autosuggestions play a central role in human cognitive regulation, shaping motivation, behavior, and decision-making. Motivated by this distinction, we simulate the effects of autosuggestions in LLMs by introducing 36 Pseudo-Cognitive Autosuggestion Statements (PCAS), which express the model’s self-directed cognition of tasks. These statements are integrated into the interaction via a fabricated chat history to explore whether LLMs exhibit behavior consistent with self-regulation mechanisms observed in humans. We conducted extensive experimental evaluations across 46 tasks using five models, including both closed- and open-source LLMs. Results show that this simulated cognitive approach systematically modulates model response strategies across contexts, giving rise to a form of pseudo-self-regulation that facilitates alignment between self-efficacy signals and actual performance. Our work offers a novel perspective on leveraging controlled pseudo-cognition to guide and regulate the behavior of language models.

Significant improvements have been observed in the zero-shot capabilities of the Large Language Models (LLMs). Due to their high sensitivity to input, research has increasingly focused on enhancing LLMs' performance via direct and simple prompt engineering rather than intricate domain adaptation. Studies suggest that LLMs exhibit emotional intelligence, and both positive and negative emotions can potentially enhance task performances. However, prior interaction prompts have predominantly concentrated on a single stimulus type, neglecting to compare different stimulus effects, examine the influence of varying task difficulties, or explore underlying mechanisms. This paper, inspired by the positive correlation between self-efficacy and task performance within the social cognitive theory, introduces Verbal Efficacy Stimulations (VES). Our VES comprises three types of verbal prompts: encouraging, provocative, and critical, addressing six aspects such as helpfulness and competence. And we further categorize task difficulty, aiming to extensively investigate how distinct VES influence the self-efficacy and task achievements of language models at varied levels of difficulty. The experimental results show that the three types of VES improve the performance of LLMs on most tasks, and the most effective VES varies for different models. In extensive experiments, we have obtained some findings consistent with psychological theories, providing novel insights for future research.

Compared to the prosperity of review domain with high-quality data for robust model evaluation, datasets from news domain are relatively scarce, and each dedicates to singular news subdomains for the Targeted Sentiment Analysis (TSA) task. This limitation hinders cross-domain evaluation, particularly for long-form Chinese news. Additionally, conventional TSA datasets are too brief, leading to the neglect of possible changes of target sentiment in lengthy texts. To address this gap, we propose a scheme to annotate sentiments towards targets in a quantitative way from a full-text perspective. Then, we introduce CNTSenti, a long-form Chinese news evaluation dataset, comprising 2,589 articles across five subfields, with an average length of 1,172 words. In addition, a domain adaptation strategy is presented to enhance the transfer of features across domains, incorporating target-guided windows, a prompt-based sentiment distribution alignment loss function, and a feature transferring mechanism utilizing contrastive learning. Extensive experiments have demonstrated the effectiveness of our approach and the challenging nature of the CNTSenti dataset.

Event sentence coreference identification (ESCI) is a fundamental task of news event detection and tracking which aims to group sentences according to events they refer to. Most recent efforts address this task by means of identifying coreferential event sentence pairs. Currently, frameworks based on pre-trained language models like Sentence-BERT (SBERT) are widely used for sentence pair tasks. However, SBERT lacks keyword awareness, while the local features of sentences can demonstrate a strong correlation with the event topic. In addition, the strategy of encoding the whole sentence is less flexible and more time-consuming. After reconsidering the significance of keywords in ESCI task, we propose KeyML, a simple keyword-based metric learning approach which leverages both lexical and semantic features of keywords to capture subject patterns of events. Specifically, a Siamese network is adapted to optimize distance metrics of keyword embeddings, resulting in more separable similarity of event sentence pairs. Then, KeyML considers keywords of data with different granularity and exploits three training strategies, along with their corresponding sampling methods, to investigate co-occurrence relationships. Experimental results show that KeyML outperforms SBERT and SimCSE on three datasets and demonstrate the effectiveness and rationality of our method.

Large Language Models (LLMs), despite their great power in language generation, often encounter challenges when dealing with intricate and knowledge-demanding queries in specific domains. This paper introduces a novel approach to enhance LLMs by effectively extracting the relevant knowledge from domain-specific textual sources, and the adaptive training of a chatbot with domain-specific inquiries. Our two-step approach starts from training a knowledge miner, namely LLMiner, which autonomously extracts Question-Answer pairs from relevant documents through a chain-of-thought reasoning process. Subsequently, we blend the mined QA pairs with a conversational dataset to fine-tune the LLM as a chatbot, thereby enriching its domain-specific expertise and conversational capabilities. We also developed a new evaluation benchmark which comprises four domain-specific text corpora and associated human-crafted QA pairs for testing. Our model shows remarkable performance improvement over generally aligned LLM and surpasses domain-adapted models directly fine-tuned on domain corpus. In particular, LLMiner achieves this with minimal human intervention, requiring only 600 seed instances, thereby providing a pathway towards self-improvement of LLMs through model-synthesized training data.

In some few-shot text classification tasks with strong data privacy or difficult labeling, the performance of pipeline methods, which directly encode text features and perform linear classification, is limited by the feature extraction ability of models. An increasing number of studies have recognized the significance of combining text features with label semantics and achieved good results. However, these existing methods cannot be well generalized to classification tasks where the class names have weak correlations with the instance texts. In this work, we address this problem by means of an effective fusion of text-label similarity and a redesign of contrastive loss. Firstly, the semantic similarity modules of text-text and text-label are adopted for further merging to improve the feature extraction ability. Then, we introduce DLSC, an inter-class differences and label semantics contrastive loss that facilitates instance embeddings to approximate correct label semantics in vector space. Experimental results show that our approach has greatly improved F1 scores on English and Chinese datasets from six classification tasks, even in tasks where label names are not strongly correlated with texts.

The release of BERT revolutionized the development of NLP. Various BERT-based reading comprehension models have been proposed, thus updating the performance ranking of reading comprehension tasks. However, the above BERT-based models inherently employ BERT’s combined input method, representing the input question and paragraph as a single packed sequence, without further modification for reading comprehension. This paper makes an in-depth analysis of this input method, proposes a problem of this approach. We call it attention deconcentration. Accordingly, this paper proposes ForceReader, a BERT-based interactive machine reading comprehension model. First, ForceReader proposes a novel solution called the Attention Separation Representation to respond to attention deconcentration. Moreover, starting from the logical nature of reading comprehension tasks, ForceReader adopts Multi-mode Reading and Interactive Reasoning strategy. For the calculation of attention, ForceReader employs Conditional Background Attention to solve the lack of the overall context semantic after the separation of attention. As an integral model, ForceReader shows a significant improvement in reading comprehension tasks compared to BERT. Moreover, this paper makes detailed visual analyses of the attention and propose strategies accordingly. This may be another argument to the explanations of the attention.

In this paper, we study how to leverage pre-trained language models in Text-to-SQL. We argue that previous approaches under utilize the base language models by concatenating all columns together with the NL question and feeding them into the base language model in the encoding stage. We propose a neat approach called Hybrid Ranking Network (HydraNet) which breaks down the problem into column-wise ranking and decoding and finally assembles the column-wise outputs into a SQL query by straightforward rules. In this approach, the encoder is given a NL question and one individual column, which perfectly aligns with the original tasks BERT/RoBERTa is trained on, and hence we avoid any ad-hoc pooling or additional encoding layers which are necessary in prior approaches. Experiments on the WikiSQL dataset show that the proposed approach is very effective, achieving the top place on the leaderboard.

Conversational virtual assistance for delivering relevant query solutions is provided. A virtual assistant system comprises various components associated with developing a knowledge database that can be searched for finding documents that fulfill the user's intent. The virtual assistant system further comprises components for receiving a query from a user, extracting entities for understanding the user's intent, and for searching a knowledge database for documents responsive to the query. When additional information is needed for determining more relevant results, a conversation strategy is determined, and a question is formulated for generating a conversation with the user for clarifying the user's intent, confirming a solution, or obtaining additional information. The user is enabled to provide a follow-up response that is related to a previously identified entity. The entity is edited in the query, and responses are refined responsive to the edited query.

The invention provides a data annotation method. The method comprises the step of data annotation task allocation, wherein according to a data identification code of to-be-annotated data and an identification code of an annotator, a to-be-annotated data annotation task is matched with the annotator, and according to a matching result, the to-be-annotated data annotation task is allocated to the annotator; the step of data annotation, wherein according to the required annotation form, the to-be-annotated data is annotated; the step of collection and integration, wherein after the annotation results of the to-be-annotated data annotation task are all submitted, according to the annotation scores of the annotator and the annotation results, the annotation result is integrated, and an accuratelabel is doped out.

The application relates to relation identification using information extracted from documents. Some implementations provide a technology and a device for mining relation information from various documents. For example, in some implementations, structured data including tables can be received. A determination that a first part of a table comprises data of a first type and a second part of the table comprises data of a second type can be made. The relation between a first content of the first part of the table and a second content of the second part of the table can be determined. The relation between the first content of the first part of the table and the second content of the second part of the table can be ranked according to a proximity and can be stored so as to create a stored relation. The stored relation can be searched based on one or more search items. A searching result of the searching based on the stored relation can be displayed. Searching results can be ranked according to a ranking associated with each stored relation.

For most entity disambiguation systems, the secret recipes are feature representations for mentions and entities, most of which are based on Bag-of-Words (BoW) representations.Commonly, BoW has several drawbacks: (1) It ignores the intrinsic meaning of words/entities; (2) It often results in high-dimension vector spaces and expensive computation; (3) For different applications, methods of designing handcrafted representations may be quite different, lacking of a general guideline.In this paper, we propose a different approach named EDKate.We first learn low-dimensional continuous vector representations for entities and words by jointly embedding knowledge base and text in the same vector space.Then we utilize these embeddings to design simple but effective features and build a two-layer disambiguation model.Extensive experiments on real-world data sets show that (1) The embedding-based features are very effective.Even a single one embedding-based feature can beat the combination of several BoW-based features.(2) The superiority is even more promising in a difficult set where the mention-entity prior cannot work well.(3) The proposed embedding method is much better than trivial implementations of some off-the-shelf embedding algorithms.(4) We compared our EDKate with existing methods/systems and the results are also positive.

Project information extraction implementations are presented that generally extract project information and generate a project information database for an enterprise. This is accomplished by extracting enterprise project names from information sources associated with an enterprise. People associated with the project corresponding to each extracted enterprise project name are identified using information sources associated with the enterprise. In addition, project-related items generated and collected during the course of the project can be identified in the information sources. A project information database is then generated for the enterprise. This database has an entry for each project which includes the extracted enterprise project name associated with the project, a list of the people identified as being associated with the project, and the project-related items or links thereto.

This paper proposes a new approach for discriminative clustering. The intuition is, for a good clustering, one should be able to learn a classifier from the clustering labels with high generalization accuracy. Thus we define a novel metric to evaluate the quality of a clustering labeling, named Minimum Separation Probability (MSP), which is a lower bound of the generalization accuracy of a classifier learnt from the clustering labeling. We take MSP as the objective to maximize and propose our approach Maximin Separation Probability Clustering (MSPC), which has several attractive properties, such as invariance to anisotropic feature scaling and intuitive probabilistic explanation for clustering quality. We present three efficient optimization strategies for MSPC, and analyze their interesting connections to existing clustering approaches, such as maximum margin clustering (MMC) and discriminative k-means. Empirical results on real world data sets verify that MSP is a robust and effective clustering quality measure. It is also shown that the proposed algorithms compare favorably to state-of-the-art clustering algorithms in both accuracy and efficiency.

We study the problem of jointly embedding a knowledge base and a text corpus.The key issue is the alignment model making sure the vectors of entities, relations and words are in the same space.Wang et al. (2014a) rely on Wikipedia anchors, making the applicable scope quite limited.In this paper we propose a new alignment model based on text descriptions of entities, without dependency on anchors.We require the embedding vector of an entity not only to fit the structured constraints in KBs but also to be equal to the embedding vector computed from the text description.Extensive experiments show that, the proposed approach consistently performs comparably or even better than the method of Wang et al. (2014a), which is encouraging as we do not use any anchor information.

Representing discrete words in a continuous vector space turns out to be useful for natural language applications related to text understanding. Meanwhile, it poses extensive challenges, one of which is due to the polysemous nature of human language. A common solution (a.k.a word sense induction) is to separate each word into multiple senses and create a representation for each sense respectively. However, this approach is usually computationally expensive and prone to data sparsity, since each sense needs to be managed discriminatively. In this work, we propose a new framework for generating context-aware text representations without diving into the sense space. We model the concept space shared among senses, resulting in a framework that is efficient in both computation and storage. Specifically, the framework we propose is one that: i) projects both words and concepts into the same vector space; ii) obtains unambiguous word representations that not only preserve the uniqueness among words, but also reflect their context-appropriate meanings. We demonstrate the effectiveness of the framework in a number of tasks on text understanding, including word/phrase similarity measurements, paraphrase identification and question-answer relatedness classification.