首页 | 本学科首页   官方微博 | 高级检索  
相似文献
 共查询到20条相似文献，搜索用时 31 毫秒
1.
Automatic text classification is the problem of automatically assigning predefined categories to free text documents, thus allowing for less manual labors required by traditional classification methods. When we apply binary classification to multi-class classification for text classification, we usually use the one-against-the-rest method. In this method, if a document belongs to a particular category, the document is regarded as a positive example of that category; otherwise, the document is regarded as a negative example. Finally, each category has a positive data set and a negative data set. But, this one-against-the-rest method has a problem. That is, the documents of a negative data set are not labeled manually, while those of a positive set are labeled by human. Therefore, the negative data set probably includes a lot of noisy data. In this paper, we propose that the sliding window technique and the revised EM (Expectation Maximization) algorithm are applied to binary text classification for solving this problem. As a result, we can improve binary text classification through extracting potentially noisy documents from the negative data set using the sliding window technique and removing actually noisy documents using the revised EM algorithm. The results of our experiments showed that our method achieved better performance than the original one-against-the-rest method in all the data sets and all the classifiers used in the experiments.  相似文献   

2.
Document classification, with the blooming of the Internet information delivery, has become indispensable required and is expected to be disposed by an automatic text categorization. This paper presents a text categorization system to solve the multi-class categorization problem. The system consists of two modules: the processing module and the classifying module. In the first module, ICF and Uni are used as the indictors to extract the relevant terms. While the fuzzy set theory is incorporated into the OAA-SVM in the classifying module, we specifically propose an OAA-FSVM classifier to implement a multi-class classification system. The performances of OAA-SVM and OAA-FSVM are evaluated by macro-average performance index.  相似文献   

3.
Diversification of web search results aims to promote documents with diverse content (i.e., covering different aspects of a query) to the top-ranked positions, to satisfy more users, enhance fairness and reduce bias. In this work, we focus on the explicit diversification methods, which assume that the query aspects are known at the diversification time, and leverage supervised learning methods to improve their performance in three different frameworks with different features and goals. First, in the LTRDiv framework, we focus on applying typical learning to rank (LTR) algorithms to obtain a ranking where each top-ranked document covers as many aspects as possible. We argue that such rankings optimize various diversification metrics (under certain assumptions), and hence, are likely to achieve diversity in practice. Second, in the AspectRanker framework, we apply LTR for ranking the aspects of a query with the goal of more accurately setting the aspect importance values for diversification. As features, we exploit several pre- and post-retrieval query performance predictors (QPPs) to estimate how well a given aspect is covered among the candidate documents. Finally, in the LmDiv framework, we cast the diversification problem into an alternative fusion task, namely, the supervised merging of rankings per query aspect. We again use QPPs computed over the candidate set for each aspect, and optimize an objective function that is tailored for the diversification goal. We conduct thorough comparative experiments using both the basic systems (based on the well-known BM25 matching function) and the best-performing systems (with more sophisticated retrieval methods) from previous TREC campaigns. Our findings reveal that the proposed frameworks, especially AspectRanker and LmDiv, outperform both non-diversified rankings and two strong diversification baselines (i.e., xQuAD and its variant) in terms of various effectiveness metrics.  相似文献   

4.
This paper presents a systematic analysis of twenty four performance measures used in the complete spectrum of Machine Learning classification tasks, i.e., binary, multi-class, multi-labelled, and hierarchical. For each classification task, the study relates a set of changes in a confusion matrix to specific characteristics of data. Then the analysis concentrates on the type of changes to a confusion matrix that do not change a measure, therefore, preserve a classifier’s evaluation (measure invariance). The result is the measure invariance taxonomy with respect to all relevant label distribution changes in a classification problem. This formal analysis is supported by examples of applications where invariance properties of measures lead to a more reliable evaluation of classifiers. Text classification supplements the discussion with several case studies.  相似文献   

5.
Artificial intelligence (AI) is rapidly becoming the pivotal solution to support critical judgments in many life-changing decisions. In fact, a biased AI tool can be particularly harmful since these systems can contribute to or demote people’s well-being. Consequently, government regulations are introducing specific rules to prohibit the use of sensitive features (e.g., gender, race, religion) in the algorithm’s decision-making process to avoid unfair outcomes. Unfortunately, such restrictions may not be sufficient to protect people from unfair decisions as algorithms can still behave in a discriminatory manner. Indeed, even when sensitive features are omitted (fairness through unawareness), they could be somehow related to other features, named proxy features. This study shows how to unveil whether a black-box model, complying with the regulations, is still biased or not. We propose an end-to-end bias detection approach exploiting a counterfactual reasoning module and an external classifier for sensitive features. In detail, the counterfactual analysis finds the minimum cost variations that grant a positive outcome, while the classifier detects non-linear patterns of non-sensitive features that proxy sensitive characteristics. The experimental evaluation reveals the proposed method’s efficacy in detecting classifiers that learn from proxy features. We also scrutinize the impact of state-of-the-art debiasing algorithms in alleviating the proxy feature problem.  相似文献   

6.
Text categorization is an important research area and has been receiving much attention due to the growth of the on-line information and of Internet. Automated text categorization is generally cast as a multi-class classification problem. Much of previous work focused on binary document classification problems. Support vector machines (SVMs) excel in binary classification, but the elegant theory behind large-margin hyperplane cannot be easily extended to multi-class text classification. In addition, the training time and scaling are also important concerns. On the other hand, other techniques naturally extensible to handle multi-class classification are generally not as accurate as SVM. This paper presents a simple and efficient solution to multi-class text categorization. Classification problems are first formulated as optimization via discriminant analysis. Text categorization is then cast as the problem of finding coordinate transformations that reflects the inherent similarity from the data. While most of the previous approaches decompose a multi-class classification problem into multiple independent binary classification tasks, the proposed approach enables direct multi-class classification. By using generalized singular value decomposition (GSVD), a coordinate transformation that reflects the inherent class structure indicated by the generalized singular values is identified. Extensive experiments demonstrate the efficiency and effectiveness of the proposed approach.  相似文献   

7.
With the increasing popularity and social influence of search engines in IR, various studies have raised concerns on the presence of bias in search engines and the social responsibilities of IR systems. As an essential component of search engine, ranking is a crucial mechanism in presenting the search results or recommending items in a fair fashion. In this article, we focus on the top-k diversity fairness ranking in terms of statistical parity fairness and disparate impact fairness. The former fairness definition provides a balanced overview of search results where the number of documents from different groups are equal; The latter enables a realistic overview where the proportion of documents from different groups reflect the overall proportion. Using 100 queries and top 100 results per query from Google as the data, we first demonstrate how topical diversity bias is present in the top web search results. Then, with our proposed entropy-based metrics for measuring the degree of bias, we reveal that the top search results are unbalanced and disproportionate to their overall diversity distribution. We explore several fairness ranking strategies to investigate the relationship between fairness, diversity, novelty and relevance. Our experimental results show that using a variant of fair ε-greedy strategy, we could bring more fairness and enhance diversity in search results without a cost of relevance. In fact, we can improve the relevance and diversity by introducing the diversity fairness. Additional experiments with TREC datasets containing 50 queries demonstrate the robustness of our proposed strategies and our findings on the impact of fairness. We present a series of correlation analysis on the amount of fairness and diversity, showing that statistical parity fairness highly correlates with diversity while disparate impact fairness does not. This provides clear and tangible implications for future works where one would want to balance fairness, diversity and relevance in search results.  相似文献   

8.
The paper is concerned with similarity search at large scale, which efficiently and effectively finds similar data points for a query data point. An efficient way to accelerate similarity search is to learn hash functions. The existing approaches for learning hash functions aim to obtain low values of Hamming distances for the similar pairs. However, these methods ignore the ranking order of these Hamming distances. This leads to the poor accuracy about finding similar items for a query data point. In this paper, an algorithm is proposed, referred to top k RHS (Rank Hash Similarity), in which a ranking loss function is designed for learning a hash function. The hash function is hypothesized to be made up of l binary classifiers. The issue of learning a hash function can be formulated as a task of learning l binary classifiers. The algorithm runs l rounds and learns a binary classifier at each round. Compared with the existing approaches, the proposed method has the same order of computational complexity. Nevertheless, experiment results on three text datasets show that the proposed method obtains higher accuracy than the baselines.  相似文献   

9.
In spite of the vast amount of work on subjectivity and sentiment analysis (SSA), it is not yet particularly clear how lexical information can best be modeled in a morphologically-richness language. To bridge this gap, we report successful models targeting lexical input in Arabic, a language of very complex morphology. Namely, we measure the impact of both gold and automatic segmentation on the task and build effective models achieving significantly higher than our baselines. Our models exploiting predicted segments improve subjectivity classification by 6.02% F1-measure and sentiment classification by 4.50% F1-measure against the majority class baseline surface word forms. We also perform in-depth (error) analyses of the behavior of the models and provide detailed explanations of subjectivity and sentiment expression in Arabic against the morphological richness background in which the work is situated.  相似文献   

10.
Real-world datasets often present different types of data quality problems, such as the presence of outliers, missing values, inaccurate representations and duplicate entities. In order to identify duplicate entities, a task named Entity Resolution (ER), we may employ a variety of classification techniques. Rule-based techniques for classification have gained increasing attention from the state of the art due to the possibility of incorporating automatic learning approaches for generating Rule-Based Entity Resolution (RbER) algorithms. However, these algorithms present a series of drawbacks: i) The generation of high-quality RbER algorithms usually require high computational and/or manual labeling costs; ii) the impossibility of tuning RbER algorithm parameters; iii) the inability to incorporate user preferences regarding the ER results in the algorithm functioning; and iv) the logical (binary) nature of the RbER algorithms usually fall short when tackling special cases, i.e., challenging duplicate and non-duplicate pairs of entities. To overcome these drawbacks, we propose Rule Assembler, a configurable approach that classifies duplicate entities based on confidence scores produced by logical rules, taking into account tunable parameters as well as user preferences. Experiments carried out using both real-world and synthetic datasets have demonstrated the ability of the proposed approach to enhance the results produced by baseline RbER algorithms and basic assembling approaches. Furthermore, we demonstrate that the proposed approach does not entail a significant overhead over the classification step and conclude that the Rule Assembler parameters APA, WPA, TβM and Max are more suitable to be used in practical scenarios.  相似文献   

11.
The aim in multi-label text classification is to assign a set of labels to a given document. Previous classifier-chain and sequence-to-sequence models have been shown to have a powerful ability to capture label correlations. However, they rely heavily on the label order, while labels in multi-label data are essentially an unordered set. The performance of these approaches is therefore highly variable depending on the order in which the labels are arranged. To avoid being dependent on label order, we design a reasoning-based algorithm named Multi-Label Reasoner (ML-Reasoner) for multi-label classification. ML-Reasoner employs a binary classifier to predict all labels simultaneously and applies a novel iterative reasoning mechanism to effectively utilize the inter-label information, where each instance of reasoning takes the previously predicted likelihoods for all labels as additional input. This approach is able to utilize information between labels, while avoiding the issue of label-order sensitivity. Extensive experiments demonstrate that our method outperforms state-of-the art approaches on the challenging AAPD dataset. We also apply our reasoning module to a variety of strong neural-based base models and show that it is able to boost performance significantly in each case.  相似文献   

12.
This paper focuses on binary optimal control of fed-batch fermentation of glycerol by Klebsiella pneumoniaewith pH feedback considering limited number of switches. To maximize the concentration of 1,3-propanediol at terminal time, we propose a binary optimal control problem subjected to time-coupled combinatorial constraint with the ratio of feeding rate of glycerol to that of NaOH as control variables. Based on time-scaling transformation and discretization, the binary optimal control problem is first transformed into a mixed binary parameter optimization problem consisting of not only continuous variables but also binary variables, which is then divided into two subproblems via combinatorial integral approximation decomposition. Finally, a novel fruit fly optimizer with modified sine cosine algorithm and adaptive maximum dwell rounding are applied to solve the obtained subproblems numerically. Numerical results show the rationality and feasibility of the proposed method.  相似文献   

13.
Influence maximization (IM) has shown wide applicability in immense fields over the past decades. Previous researches on IM mainly focused on the dyadic relationship but lacked the consideration of higher-order relationship between entities, which has been constantly revealed in many real systems. An adaptive degree-based heuristic algorithm, i.e., Hyper Adaptive Degree Pruning (HADP) which aims to iteratively select nodes with low influence overlap as seeds, is proposed in this work to tackle the IM problem in hypergraphs. Furthermore, we extend algorithms from ordinary networks as baselines. Results on 8 empirical hypergraphs show that HADP surpasses the baselines in terms of both effectiveness and efficiency with a maximally 46.02% improvement. Moreover, we test the effectiveness of our algorithm on synthetic hypergraphs generated by different degree heterogeneity. It shows that the improvement of our algorithm effectiveness increases from 2.66% to 14.67% with the increase of degree heterogeneity, which indicates that HADP shows high performance especially in hypergraphs with high heterogeneity, which is ubiquitous in real-world systems.  相似文献   

14.
Researchers in indexing and retrieval systems have been advocating the inclusion of more contextual information to improve results. The proliferation of full-text databases and advances in computer storage capacity have made it possible to carry out text analysis by means of linguistic and extra-linguistic knowledge. Since the mid 80s, research has tended to pay more attention to context, giving discourse analysis a more central role. The research presented in this paper aims to check whether discourse variables have an impact on modern information retrieval and classification algorithms. In order to evaluate this hypothesis, a functional framework for information analysis in an automated environment has been proposed, where the n-grams (filtering) and the k-means and Chen’s classification algorithms have been tested against sub-collections of documents based on the following discourse variables: “Genre”, “Register”, “Domain terminology”, and “Document structure”. The results obtained with the algorithms for the different sub-collections were compared to the MeSH information structure. These demonstrate that n-grams does not appear to have a clear dependence on discourse variables, though the k-means classification algorithm does, but only on domain terminology and document structure, and finally Chen’s algorithm has a clear dependence on all of the discourse variables. This information could be used to design better classification algorithms, where discourse variables should be taken into account. Other minor conclusions drawn from these results are also presented.  相似文献   

15.
Learning low dimensional dense representations of the vocabularies of a corpus, known as neural embeddings, has gained much attention in the information retrieval community. While there have been several successful attempts at integrating embeddings within the ad hoc document retrieval task, yet, no systematic study has been reported that explores the various aspects of neural embeddings and how they impact retrieval performance. In this paper, we perform a methodical study on how neural embeddings influence the ad hoc document retrieval task. More specifically, we systematically explore the following research questions: (i) do methods solely based on neural embeddings perform competitively with state of the art retrieval methods with and without interpolation? (ii) are there any statistically significant difference between the performance of retrieval models when based on word embeddings compared to when knowledge graph entity embeddings are used? and (iii) is there significant difference between using locally trained neural embeddings compared to when globally trained neural embeddings are used? We examine these three research questions across both hard and all queries. Our study finds that word embeddings do not show competitive performance to any of the baselines. In contrast, entity embeddings show competitive performance to the baselines and when interpolated, outperform the best baselines for both hard and soft queries.  相似文献   

16.
Representation learning has recently been used to remove sensitive information from data and improve the fairness of machine learning algorithms in social applications. However, previous works that used neural networks are opaque and poorly interpretable, as it is difficult to intuitively determine the independence between representations and sensitive information. The internal correlation among data features has not been fully discussed, and it may be the key to improving the interpretability of neural networks. A novel fair representation algorithm referred to as FRC is proposed from this conjecture. It indicates how representations independent of multiple sensitive attributes can be learned by applying specific correlation constraints on representation dimensions. Specifically, dimensions of the representation and sensitive attributes are treated as statistical variables. The representation variables are divided into two parts related to and unrelated to the sensitive variables by adjusting their absolute correlation coefficient with sensitive variables. The potential impact of sensitive information on representations is concentrated in the related part. The unrelated part of the representation can be used in downstream tasks to yield fair results. FRC takes the correlation between dimensions as the key to solving the problem of fair representation. Empirical results show that our representations enhance the ability of neural networks to show fairness and achieve better fairness-accuracy tradeoffs than state-of-the-art works.  相似文献   

17.
Event relations specify how different event flows expressed within the context of a textual passage relate to each other in terms of temporal and causal sequences. There have already been impactful work in the area of temporal and causal event relation extraction; however, the challenge with these approaches is that (1) they are mostly supervised methods and (2) they rely on syntactic and grammatical structure patterns at the sentence-level. In this paper, we address these challenges by proposing an unsupervised event network representation for temporal and causal relation extraction that operates at the document level. More specifically, we benefit from existing Open IE systems to generate a set of triple relations that are then used to build an event network. The event network is bootstrapped by labeling the temporal disposition of events that are directly linked to each other. We then systematically traverse the event network to identify the temporal and causal relations between indirectly connected events. We perform experiments based on the widely adopted TempEval-3 and Causal-TimeBank corpora and compare our work with several strong baselines. We show that our method improves performance compared to several strong methods.  相似文献   

18.
This paper is concerned with the distributed H∞ filtering problem for a class of sensor networks with stochastic sampling. System measurements are collected through a sensor network stochastically and the phenomena such as random measurement missing and quantization are also considered. Firstly, the stochastic sampling process of the sensor network is modeled as a discrete-time Markovian system. Then, the logarithmic quantization effect is transformed into the parameter uncertainty of the filtering system, and a set of binary variables is introduced to model the random measurement missing phenomenon. Finally, the resulting augmented system is modeled as an uncertain Markovian system with multiple random variables. Based on the Lyapunov stability theory and the stochastic system analysis method, a sufficient condition is obtained such that the augmented system is stochastically stable and achieves an average H∞ performance level γ; the design procedure of the optimal distributed filter is also provided. A numerical example is given to demonstrate the effectiveness of the proposed results.  相似文献   

19.
This paper compares 14 information retrieval metrics based on graded relevance, together with 10 traditional metrics based on binary relevance, in terms of stability, sensitivity and resemblance of system rankings. More specifically, we compare these metrics using the Buckley/Voorhees stability method, the Voorhees/Buckley swap method and Kendall’s rank correlation, with three data sets comprising test collections and submitted runs from NTCIR. Our experiments show that (Average) Normalised Discounted Cumulative Gain at document cut-off l are the best among the rank-based graded-relevance metrics, provided that l is large. On the other hand, if one requires a recall-based graded-relevance metric that is highly correlated with Average Precision, then Q-measure is the best choice. Moreover, these best graded-relevance metrics are at least as stable and sensitive as Average Precision, and are fairly robust to the choice of gain values.  相似文献   

20.
In this paper we demonstrate a new method for concentrating the set of key-words of a thesaurus. This method is based on a mathematical study that we have carried out into the distribution of characters in a defined natural language.We have built a function f of concentration which generates only a few synonyms. In applying this function to the set of key-words of a thesaurus, we reduce each key-word to four characters without synonymity. (For three characters we have a rate of synonymity of approx. 1/1000th.)A new structure of binary files allows the thesaurus to be contained in a table of less than 700 bytes.  相似文献   

设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号