Search (24 results, page 1 of 2)

0.03406783 = product of:
  0.06813566 = sum of:
    0.042235587 = weight(_text_:data in 2748) [ClassicSimilarity], result of:
      0.042235587 = score(doc=2748,freq=2.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.34936053 = fieldWeight in 2748, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.078125 = fieldNorm(doc=2748)
    0.02590007 = product of:
      0.05180014 = sum of:
        0.05180014 = weight(_text_:22 in 2748) [ClassicSimilarity], result of:
          0.05180014 = score(doc=2748,freq=2.0), product of:
            0.13388468 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.03823278 = queryNorm
            0.38690117 = fieldWeight in 2748, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.078125 = fieldNorm(doc=2748)
      0.5 = coord(1/2)
  0.5 = coord(2/4)

Date

1. 2.2016 18:25:22

Source

Semantic keyword-based search on structured data sources: First COST Action IC1302 International KEYSTONE Conference, IKC 2015, Coimbra, Portugal, September 8-9, 2015. Revised Selected Papers. Eds.: J. Cardoso et al

0.011946027 = product of:
  0.04778411 = sum of:
    0.04778411 = weight(_text_:data in 4095) [ClassicSimilarity], result of:
      0.04778411 = score(doc=4095,freq=16.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.3952563 = fieldWeight in 4095, product of:
          4.0 = tf(freq=16.0), with freq of:
            16.0 = termFreq=16.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03125 = fieldNorm(doc=4095)
  0.25 = coord(1/4)

Abstract

In traditional text classification, classes are mutually exclusive, i.e. it is not possible to have one text or text fragment classified into more than one class. On the other hand, in multi-label classification an individual text may belong to several classes simultaneously. This type of classification is required by a large number of current applications such as big data classification, images and video annotation. Supervised learning is the most used type of machine learning in the classification task. It requires large quantities of labeled data and the intervention of a human tagger in the creation of the training sets. When the data sets become very large or heavily noisy, this operation can be tedious, prone to error and time consuming. In this case, semi-supervised learning, which requires only few labels, is a better choice. In this paper, we study and evaluate several methods to address the problem of multi-label classification using semi-supervised learning and data from social networks. First, we propose a linguistic pre-processing involving tokeni-sation, recognition of named entities and hashtag segmentation in order to decrease the noise in this type of massive and unstructured real data and then we perform a word sense disambiguation using WordNet. Second, several experiments related to multi-label classification and semi-supervised learning are carried out on these data sets and compared to each other. These evaluations compare the results of the approaches considered. This paper proposes a method for combining semi-supervised methods with a graph method for the extraction of subjects in social networks using a multi-label classification approach. Experiments show that the performance of the proposed model increases in 4 p.p. the precision of the classification when compared to a baseline.

Source

IEEE International Conference on Big Data (Big Data) (2017)

0.011805206 = product of:
  0.047220822 = sum of:
    0.047220822 = weight(_text_:data in 3014) [ClassicSimilarity], result of:
      0.047220822 = score(doc=3014,freq=10.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.39059696 = fieldWeight in 3014, product of:
          3.1622777 = tf(freq=10.0), with freq of:
            10.0 = termFreq=10.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=3014)
  0.25 = coord(1/4)

Abstract

Automatic text classification (TC) continues to be a relevant research topic and several TC algorithms have been proposed. However, the majority of TC algorithms assume that the underlying data distribution does not change over time. In this work, we are concerned with the challenges imposed by the temporal dynamics observed in textual data sets. We provide evidence of the existence of temporal effects in three textual data sets, reflected by variations observed over time in the class distribution, in the pairwise class similarities, and in the relationships between terms and classes. We then quantify, using a series of full factorial design experiments, the impact of these effects on four well-known TC algorithms. We show that these temporal effects affect each analyzed data set differently and that they restrict the performance of each considered TC algorithm to different extents. The reported quantitative analyses, which are the original contributions of this article, provide valuable new insights to better understand the behavior of TC algorithms when faced with nonstatic (temporal) data distributions and highlight important requirements for the proposal of more accurate classification models.

0.010973128 = product of:
  0.04389251 = sum of:
    0.04389251 = weight(_text_:data in 3464) [ClassicSimilarity], result of:
      0.04389251 = score(doc=3464,freq=6.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.3630661 = fieldWeight in 3464, product of:
          2.4494898 = tf(freq=6.0), with freq of:
            6.0 = termFreq=6.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.046875 = fieldNorm(doc=3464)
  0.25 = coord(1/4)

Abstract

We propose a new hybrid clustering framework to incorporate text mining with bibliometrics in journal set analysis. The framework integrates two different approaches: clustering ensemble and kernel-fusion clustering. To improve the flexibility and the efficiency of processing large-scale data, we propose an information-based weighting scheme to leverage the effect of multiple data sources in hybrid clustering. Three different algorithms are extended by the proposed weighting scheme and they are employed on a large journal set retrieved from the Web of Science (WoS) database. The clustering performance of the proposed algorithms is systematically evaluated using multiple evaluation methods, and they were cross-compared with alternative methods. Experimental results demonstrate that the proposed weighted hybrid clustering strategy is superior to other methods in clustering performance and efficiency. The proposed approach also provides a more refined structural mapping of journal sets, which is useful for monitoring and detecting new trends in different scientific fields.

Theme

Data Mining

0.010973128 = product of:
  0.04389251 = sum of:
    0.04389251 = weight(_text_:data in 3015) [ClassicSimilarity], result of:
      0.04389251 = score(doc=3015,freq=6.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.3630661 = fieldWeight in 3015, product of:
          2.4494898 = tf(freq=6.0), with freq of:
            6.0 = termFreq=6.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.046875 = fieldNorm(doc=3015)
  0.25 = coord(1/4)

Abstract

We analyze the linguistic evolution of selected scientific disciplines over a 30-year time span (1970s to 2000s). Our focus is on four highly specialized disciplines at the boundaries of computer science that emerged during that time: computational linguistics, bioinformatics, digital construction, and microelectronics. Our analysis is driven by the question whether these disciplines develop a distinctive language use-both individually and collectively-over the given time period. The data set is the English Scientific Text Corpus (scitex), which includes texts from the 1970s/1980s and early 2000s. Our theoretical basis is register theory. In terms of methods, we combine corpus-based methods of feature extraction (various aggregated features [part-of-speech based], n-grams, lexico-grammatical patterns) and automatic text classification. The results of our research are directly relevant to the study of linguistic variation and languages for specific purposes (LSP) and have implications for various natural language processing (NLP) tasks, for example, authorship attribution, text mining, or training NLP tools.

Theme

Data Mining

0.010558897 = product of:
  0.042235587 = sum of:
    0.042235587 = weight(_text_:data in 5055) [ClassicSimilarity], result of:
      0.042235587 = score(doc=5055,freq=8.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.34936053 = fieldWeight in 5055, product of:
          2.828427 = tf(freq=8.0), with freq of:
            8.0 = termFreq=8.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=5055)
  0.25 = coord(1/4)

Abstract

Distant supervision (DS) has the advantage of automatically generating large amounts of labelled training data and has been widely used for relation extraction. However, there are usually many wrong labels in the automatically labelled data in distant supervision (Riedel, Yao, & McCallum, 2010). This paper presents a novel method to reduce the wrong labels. The proposed method uses the semantic Jaccard with word embedding to measure the semantic similarity between the relation phrase in the knowledge base and the dependency phrases between two entities in a sentence to filter the wrong labels. In the process of reducing wrong labels, the semantic Jaccard algorithm selects a core dependency phrase to represent the candidate relation in a sentence, which can capture features for relation classification and avoid the negative impact from irrelevant term sequences that previous neural network models of relation extraction often suffer. In the process of relation classification, the core dependency phrases are also used as the input of a convolutional neural network (CNN) for relation classification. The experimental results show that compared with the methods using original DS data, the methods using filtered DS data performed much better in relation extraction. It indicates that the semantic similarity based method is effective in reducing wrong labels. The relation extraction performance of the CNN model using the core dependency phrases as input is the best of all, which indicates that using the core dependency phrases as input of CNN is enough to capture the features for relation classification and could avoid negative impact from irrelevant terms.

0.009144273 = product of:
  0.03657709 = sum of:
    0.03657709 = weight(_text_:data in 2706) [ClassicSimilarity], result of:
      0.03657709 = score(doc=2706,freq=6.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.30255508 = fieldWeight in 2706, product of:
          2.4494898 = tf(freq=6.0), with freq of:
            6.0 = termFreq=6.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=2706)
  0.25 = coord(1/4)

Abstract

Transfer learning utilizes labeled data available from some related domain (source domain) for achieving effective knowledge transformation to the target domain. However, most state-of-the-art cross-domain classification methods treat documents as plain text and ignore the hyperlink (or citation) relationship existing among the documents. In this paper, we propose a novel cross-domain document classification approach called Link-Bridged Topic model (LBT). LBT consists of two key steps. Firstly, LBT utilizes an auxiliary link network to discover the direct or indirect co-citation relationship among documents by embedding the background knowledge into a graph kernel. The mined co-citation relationship is leveraged to bridge the gap across different domains. Secondly, LBT simultaneously combines the content information and link structures into a unified latent topic model. The model is based on an assumption that the documents of source and target domains share some common topics from the point of view of both content information and link structure. By mapping both domains data into the latent topic spaces, LBT encodes the knowledge about domain commonality and difference as the shared topics with associated differential probabilities. The learned latent topics must be consistent with the source and target data, as well as content and link statistics. Then the shared topics act as the bridge to facilitate knowledge transfer from the source to the target domains. Experiments on different types of datasets show that our algorithm significantly improves the generalization performance of cross-domain document classification.

0.009144273 = product of:
  0.03657709 = sum of:
    0.03657709 = weight(_text_:data in 3121) [ClassicSimilarity], result of:
      0.03657709 = score(doc=3121,freq=6.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.30255508 = fieldWeight in 3121, product of:
          2.4494898 = tf(freq=6.0), with freq of:
            6.0 = termFreq=6.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=3121)
  0.25 = coord(1/4)

Abstract

The delineation of coordinates is fundamental for the cartography of science, and accurate and credible classification of scientific knowledge presents a persistent challenge in this regard. We present a map of Finnish science based on unsupervised-learning classification, and discuss the advantages and disadvantages of this approach vis-à-vis those generated by human reasoning. We conclude that from theoretical and practical perspectives there exist several challenges for human reasoning-based classification frameworks of scientific knowledge, as they typically try to fit new-to-the-world knowledge into historical models of scientific knowledge, and cannot easily be deployed for new large-scale data sets. Automated classification schemes, in contrast, generate classification models only from the available text corpus, thereby identifying credibly novel bodies of knowledge. They also lend themselves to versatile large-scale data analysis, and enable a range of Big Data possibilities. However, we also argue that it is neither possible nor fruitful to declare one or another method a superior approach in terms of realism to classify scientific knowledge, and we believe that the merits of each approach are dependent on the practical objectives of analysis.

0.0074662673 = product of:
  0.02986507 = sum of:
    0.02986507 = weight(_text_:data in 4775) [ClassicSimilarity], result of:
      0.02986507 = score(doc=4775,freq=4.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.24703519 = fieldWeight in 4775, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=4775)
  0.25 = coord(1/4)

Abstract

The progressive increase of information content has recently made it necessary to create a system for automatic classification of documents. In this article, a system is presented for the categorization of multiclass Farsi documents that requires fewer training examples and can help to compensate the shortcoming of the standard training dataset. The new idea proposed in the present article is based on extending the feature vector by adding some words extracted from a thesaurus and then filtering the new feature vector by applying secondary feature selection to discard inappropriate features. In fact, a phase of secondary feature selection is applied to choose more appropriate features among the features added from a thesaurus to enhance the effect of using a thesaurus on the efficiency of the classifier. To evaluate the proposed system, a corpus is gathered from the Farsi Wikipedia website and some articles in the Hamshahri newspaper, the Roshd periodical, and the Soroush magazine. In addition to studying the role of a thesaurus and applying secondary feature selection, the effect of a various number of categories, size of the training dataset, and average number of words in the test data also are examined. As the results indicate, classification efficiency improves by applying this approach, especially when available data is not sufficient for some text categories.

0.0074662673 = product of:
  0.02986507 = sum of:
    0.02986507 = weight(_text_:data in 237) [ClassicSimilarity], result of:
      0.02986507 = score(doc=237,freq=4.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.24703519 = fieldWeight in 237, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=237)
  0.25 = coord(1/4)

Abstract

We study the problem of question topic classification using a very large real-world Community Question Answering (CQA) dataset from Yahoo! Answers. The dataset comprises 3.9 million questions and these questions are organized into more than 1,000 categories in a hierarchy. To the best knowledge, this is the first systematic evaluation of the performance of different classification methods on question topic classification as well as short texts. Specifically, we empirically evaluate the following in classifying questions into CQA categories: (a) the usefulness of n-gram features and bag-of-word features; (b) the performance of three standard classification algorithms (naive Bayes, maximum entropy, and support vector machines); (c) the performance of the state-of-the-art hierarchical classification algorithms; (d) the effect of training data size on performance; and (e) the effectiveness of the different components of CQA data, including subject, content, asker, and the best answer. The experimental results show what aspects are important for question topic classification in terms of both effectiveness and efficiency. We believe that the experimental findings from this study will be useful in real-world classification problems.

0.0074662673 = product of:
  0.02986507 = sum of:
    0.02986507 = weight(_text_:data in 967) [ClassicSimilarity], result of:
      0.02986507 = score(doc=967,freq=4.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.24703519 = fieldWeight in 967, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=967)
  0.25 = coord(1/4)

Abstract

Because of Twitter's popularity and the viral nature of information dissemination on Twitter, predicting which Twitter topics will become popular in the near future becomes a task of considerable economic importance. Many Twitter topics are annotated by hashtags. In this article, we propose methods to predict the popularity of new hashtags on Twitter by formulating the problem as a classification task. We use five standard classification models (i.e., Naïve bayes, k-nearest neighbors, decision trees, support vector machines, and logistic regression) for prediction. The main challenge is the identification of effective features for describing new hashtags. We extract 7 content features from a hashtag string and the collection of tweets containing the hashtag and 11 contextual features from the social graph formed by users who have adopted the hashtag. We conducted experiments on a Twitter data set consisting of 31 million tweets from 2 million Singapore-based users. The experimental results show that the standard classifiers using the extracted features significantly outperform the baseline methods that do not use these features. Among the five classifiers, the logistic regression model performs the best in terms of the Micro-F1 measure. We also observe that contextual features are more effective than content features.

Theme

Data Mining

0.0074662673 = product of:
  0.02986507 = sum of:
    0.02986507 = weight(_text_:data in 1056) [ClassicSimilarity], result of:
      0.02986507 = score(doc=1056,freq=4.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.24703519 = fieldWeight in 1056, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=1056)
  0.25 = coord(1/4)

Abstract

The ever increasing amount of digitally available information is curse and blessing at the same time. On the one hand, users have increasingly large amounts of information at their fingertips. On the other hand, the assessment and refinement of web search results becomes more and more tiresome and difficult for non-experts in a domain. Therefore, established digital libraries offer specialized collections with a certain degree of quality. This quality can largely be attributed to the great effort invested into semantic enrichment of the provided documents e.g. by annotating their documents with respect to a domain-specific taxonomy. This process is still done manually in many domains, e.g. chemistry CAS, medicine MeSH, or mathematics MSC. But due to the growing amount of data, this manual task gets more and more time consuming and expensive. The only solution for this problem seems to employ automated classification algorithms, but from evaluations done in previous research, conclusions to a real world scenario are difficult to make. We therefore conducted a large scale feasibility study on a real world data set from one of the biggest mathematical digital libraries, i.e. Zentralblatt MATH, with special focus on its practical applicability.

0.0074662673 = product of:
  0.02986507 = sum of:
    0.02986507 = weight(_text_:data in 2852) [ClassicSimilarity], result of:
      0.02986507 = score(doc=2852,freq=4.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.24703519 = fieldWeight in 2852, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=2852)
  0.25 = coord(1/4)

Abstract

Social tagging is one of the most popular methods for collecting crowd-sourced information in galleries, libraries, archives, and museums (GLAMs). However, when the number of social tags grows rapidly, using them becomes problematic and, as a result, they are often left as simply big data that cannot be used for practical purposes. To revitalize the use of this crowd-sourced information, we propose using social tags to link and cluster artworks based on an experimental study using an online collection at the Gyeonggi Museum of Modern Art (GMoMA). We view social tagging as a folksonomy, where artworks are classified by keywords of the crowd's various interpretations and one artwork can belong to several different categories simultaneously. To leverage this strength of social tags, we used a clustering method called "link communities" to detect overlapping communities in a network of artworks constructed by computing similarities between all artwork pairs. We used this framework to identify semantic relationships and clusters of similar artworks. By comparing the clustering results with curators' manual classification results, we demonstrated the potential of social tagging data for automatically clustering artworks in a way that reflects the dynamic perspectives of crowds.

0.0063353376 = product of:
  0.02534135 = sum of:
    0.02534135 = weight(_text_:data in 3331) [ClassicSimilarity], result of:
      0.02534135 = score(doc=3331,freq=2.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.2096163 = fieldWeight in 3331, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.046875 = fieldNorm(doc=3331)
  0.25 = coord(1/4)

Abstract

Automatic text classification (TC) is essential for the management of information. To properly classify a document d, it is essential to identify the semantics of each term t in d, while the semantics heavily depend on context (neighboring terms) of t in d. Therefore, we present a technique CTFA (Context-based Term Frequency Assessment) that improves text classifiers by considering term contexts in test documents. The results of the term context recognition are used to assess term frequencies of terms, and hence CTFA may easily work with various kinds of text classifiers that base their TC decisions on term frequencies, without needing to modify the classifiers. Moreover, CTFA is efficient, and neither huge memory nor domain-specific knowledge is required. Empirical results show that CTFA successfully enhances performance of several kinds of text classifiers on different experimental data.

0.0063353376 = product of:
  0.02534135 = sum of:
    0.02534135 = weight(_text_:data in 4558) [ClassicSimilarity], result of:
      0.02534135 = score(doc=4558,freq=2.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.2096163 = fieldWeight in 4558, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.046875 = fieldNorm(doc=4558)
  0.25 = coord(1/4)

Abstract

While automated methods for information organization have been around for several decades now, exponential growth of the World Wide Web has put them into the forefront of research in different communities, within which several approaches can be identified: 1) machine learning (algorithms that allow computers to improve their performance based on learning from pre-existing data); 2) document clustering (algorithms for unsupervised document organization and automated topic extraction); and 3) string matching (algorithms that match given strings within larger text). Here the aim was to automatically organize textual documents into hierarchical structures for subject browsing. The string-matching approach was tested using a controlled vocabulary (containing pre-selected and pre-defined authorized terms, each corresponding to only one concept). The results imply that an appropriate controlled vocabulary, with a sufficient number of entry terms designating classes, could in itself be a solution for automated classification. Then, if the same controlled vocabulary had an appropriat hierarchical structure, it would at the same time provide a good browsing structure for the collection of automatically classified documents.

0.0063353376 = product of:
  0.02534135 = sum of:
    0.02534135 = weight(_text_:data in 1496) [ClassicSimilarity], result of:
      0.02534135 = score(doc=1496,freq=2.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.2096163 = fieldWeight in 1496, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.046875 = fieldNorm(doc=1496)
  0.25 = coord(1/4)

Abstract

An important aspect to performing text categorization is selecting appropriate supervised classification and feature selection methods. A comprehensive benchmark is needed to inform best practices in this broad application field. Previous benchmarks have evaluated performance for a few supervised classification and feature selection methods and limited ways to optimize them. The present work updates prior benchmarks by increasing the number of classifiers and feature selection methods order of magnitude, including adding recently developed, state-of-the-art methods. Specifically, this study used 229 text categorization data sets/tasks, and evaluated 28 classification methods (both well-established and proprietary/commercial) and 19 feature selection methods according to 4 classification performance metrics. We report several key findings that will be helpful in establishing best methodological practices for text categorization.

0.0059730136 = product of:
  0.023892054 = sum of:
    0.023892054 = weight(_text_:data in 2301) [ClassicSimilarity], result of:
      0.023892054 = score(doc=2301,freq=4.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.19762816 = fieldWeight in 2301, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03125 = fieldNorm(doc=2301)
  0.25 = coord(1/4)

Abstract

Analytico-synthetic and faceted classifications, such as Universal Decimal Classification (UDC) express content of documents with complex, pre-combined classification codes. Without classification authority control that would help manage and access structured notations, the use of UDC codes in searching and browsing is limited. Existing UDC parsing solutions are usually created for a particular database system or a specific task and are not widely applicable. The approach described in this paper provides a solution by which the analysis and interpretation of UDC notations would be stored into an intermediate format (in this case, in XML) by automatic means without any data or information loss. Due to its richness, the output file can be converted into different formats, such as standard mark-up and data exchange formats or simple lists of the recommended entry points of a UDC number. The program can also be used to create authority records containing complex UDC numbers which can be comprehensively analysed in order to be retrieved effectively. The Java program, as well as the corresponding schema definition it employs, is under continuous development. The current version of the interpreter software is now available online for testing purposes at the following web site: http://interpreter-eto.rhcloud.com. The future plan is to implement conversion methods for standard formats and to create standard online interfaces in order to make it possible to use the features of software as a service. This would result in the algorithm being able to be employed both in existing and future library systems to analyse UDC numbers without any significant programming effort.

0.0052794483 = product of:
  0.021117793 = sum of:
    0.021117793 = weight(_text_:data in 472) [ClassicSimilarity], result of:
      0.021117793 = score(doc=472,freq=2.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.17468026 = fieldWeight in 472, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=472)
  0.25 = coord(1/4)

Abstract

The following paper deals with an automatic text classification method which does not require training documents. For this method the German Subject Heading Authority File (SWD), provided by the linked data service of the German National Library is used. Recently the SWD was enriched with notations of the Dewey Decimal Classification (DDC). In consequence it became possible to utilize the subject headings as textual representations for the notations of the DDC. Basically, we we derive the classification of a text from the classification of the words in the text given by the thesaurus. The method was tested by classifying 3826 OAI-Records from 7 different repositories. Mean reciprocal rank and recall were chosen as evaluation measure. Direct comparison to a machine learning method has shown that this method is definitely competitive. Thus we can conclude that the enriched version of the SWD provides high quality information with a broad coverage for classification of German scientific articles.

0.0052794483 = product of:
  0.021117793 = sum of:
    0.021117793 = weight(_text_:data in 2300) [ClassicSimilarity], result of:
      0.021117793 = score(doc=2300,freq=2.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.17468026 = fieldWeight in 2300, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=2300)
  0.25 = coord(1/4)

Abstract

Subject terms play a crucial role in resource discovery but require substantial effort to produce. Automatic subject classification and indexing address problems of scale and sustainability and can be used to enrich existing bibliographic records, establish more connections across and between resources and enhance consistency of bibliographic data. The paper aims to put forward a complex methodological framework to evaluate automatic classification tools of Swedish textual documents based on the Dewey Decimal Classification (DDC) recently introduced to Swedish libraries. Three major complementary approaches are suggested: a quality-built gold standard, retrieval effects, domain analysis. The gold standard is built based on input from at least two catalogue librarians, end-users expert in the subject, end users inexperienced in the subject and automated tools. Retrieval effects are studied through a combination of assigned and free tasks, including factual and comprehensive types. The study also takes into consideration the different role and character of subject terms in various knowledge domains, such as scientific disciplines. As a theoretical framework, domain analysis is used and applied in relation to the implementation of DDC in Swedish libraries and chosen domains of knowledge within the DDC itself.

0.0052794483 = product of:
  0.021117793 = sum of:
    0.021117793 = weight(_text_:data in 2722) [ClassicSimilarity], result of:
      0.021117793 = score(doc=2722,freq=2.0), product of:
        0.120893985 = queryWeight, product of:
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.03823278 = queryNorm
        0.17468026 = fieldWeight in 2722, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          3.1620505 = idf(docFreq=5088, maxDocs=44218)
          0.0390625 = fieldNorm(doc=2722)
  0.25 = coord(1/4)

Abstract

We present a Life-Long Learning from Mistakes (3LM) algorithm for document classification, which could be used in various scenarios such as spam filtering, blog classification, and web resource categorization. We extend the ideas of online clustering and batch-mode centroid-based classification to online learning with negative feedback. The 3LM is a competitive learning algorithm, which avoids over-smoothing, characteristic of the centroid-based classifiers, by using a different class representative, which we call clusterhead. The clusterheads competing for vector-space dominance are drawn toward misclassified documents, eventually bringing the model to a "balanced state" for a fixed distribution of documents. Subsequently, the clusterheads oscillate between the misclassified documents, heuristically minimizing the rate of misclassifications, an NP-complete problem. Further, the 3LM algorithm prevents over-fitting by "leashing" the clusterheads to their respective centroids. A clusterhead provably converges if its class can be separated by a hyper-plane from all other classes. Lifelong learning with fixed learning rate allows 3LM to adapt to possibly changing distribution of the data and continually learn and unlearn document classes. We report on our experiments, which demonstrate high accuracy of document classification on Reuters21578, OHSUMED, and TREC07p-spam datasets. The 3LM algorithm did not show over-fitting, while consistently outperforming centroid-based, Naïve Bayes, C4.5, AdaBoost, kNN, and SVM whose accuracy had been reported on the same three corpora.