Text Clustering

Data-mining techniques that detect trends and patterns in structured data are often illsuited for analysis of unstructured text. Information critical to business-and generated by groups such as employees, customers, and the public-appears in such forms as chats, electronic discussion forums, and blogs. This paper describes techniques developed to detect themes and trends in such informal communication streams. Our approach begins with unsupervised text clustering to create initial categories. A human analyst then refines the categories into easily understandable themes. To facilitate this process, we developed an interactive approach to text category creation and validation that aids the analyst in evaluating each category of a taxonomy and makes it possible to visualize relationships among categories. The resulting analysis can then be communicated to participants in real time. We report on the results of using these techniques in IBM companywide ''Jam'' events, during which tens of thousands of employees worldwide participated in electronic discussions of key business issues.

The tagging aims to address a challenge to search relevant text-documents given a set of tags. In addition, the tag-based approaches received a wide attention as a possible solution to the big-content. Probabilistic topic model methods, such as Dirichlet distribution and non-negative matrix factorization are used for tagging process. Both have many challenges. The iterations in addition the semantic coherence are considered as challenges in semantic tagging applications. In light of this, we propose a novel learning tagging model called semantic non-negative matrix factorization, which introduces the utilization of the semantic text representation via knowledge-based approach to extract the term-topic matrix and the topic-document matrix by semantically approach. The proposed words are based on a novel initialization method for non-negative matrix factorization technique. In the experimental evaluations, we use five datasets demonstrate the effectiveness of our model. The results are compared with the state-of-the-art model. The results show the proposed model has an ability to generate more precise topics with semantically related and having the high sense to the disambiguation of meaning, provides up to more dimensionality reduction and the topic coherence based semantic.

The detection and extraction of scene text from document images is one of the challenging research areas. Many researchers have detected and extracted the text from plain text background. But the multi-oriented scene text detection is one of the complex problems due to multi-oriented texts which have different orientation, font size, colors etc. In this work, we have proposed a new algorithm to detect and extract the multi-oriented scene text. Experiments have been carried out to find the robustness on the proposed method, by conducting various experiments on heterogeneous datasets. The proposed method achieves selectively high detection rate of 88.43% on the multi-oriented scene text.

Results of queries by personal names often contain documents related to several people because of the namesake problem. In order to differentiate documents related to different people, an effective method is needed to measure document similarities and to find documents related to the same person. Some previous researchers have used the vector space model or have tried to extract common named entities for measuring similarities. We propose a new method that uses Web directories as a knowledge base to find shared contexts in document pairs and uses the measurement of shared contexts to determine similarities between document pairs. Experimental results show that our proposed method outperforms the vector space model method and the named entity recognition method.

This paper summarizes the goals, organization and results of the first RepLab competitive evaluation campaign for Online Reputation Management Systems (RepLab 2012). RepLab focused on the reputation of companies, and asked participant systems to annotate different types of information on tweets containing the names of several companies. Two tasks were proposed: a profiling task, where tweets had to be annotated for relevance and polarity for reputation, and a monitoring task, where tweets had to be clustered thematically and clusters had to be ordered by priority (for reputation management purposes). The gold standard consisted of annotations made by reputation management experts, a feature which turns the RepLab 2012 test collection in a useful source not only to evaluate systems, but also to reach a better understanding of the notions of polarity and priority in the context of reputation management.

The third WePS (Web People Search) Evaluation campaign took place in 2009-2010 and attracted the participation of 13 research groups from Europe, Asia and North America. Given the top web search results for a person name, two tasks were addressed: a clustering task, which consists of grouping together web pages referring to the same person, and an extraction task, which consists of extracting salient attributes for each of the persons sharing the same name. Continuing the path of previous campaigns, this third evaluation aimed at merging both problems into one single task, where the system must return both the documents and the attributes for each of the different people sharing a given name. This is not a trivial step from the point of view of evaluation: a system may correctly extract attribute profiles from different URLs but then incorrectly merge profiles. This campaign also featured a larger testbed and the participation of a state-of-the-art commercial WePS system in the attribute extraction task. This paper presents the definition, resources, evaluation methodology and results for the clustering and attribute extraction tasks.

Text mining, also known as discovering knowledge from the text, which has emerged as a possible solution for the current information explosion, refers to the process of extracting non-trivial and useful patterns from unstructured text. Among the general tasks of text mining such as text clustering, summarization, etc, text classification is a subtask of intelligent information processing, which employs unsupervised learning to construct a classifier from training text by which to predict the class of unlabeled text. Because of its simplicity and objectivity in performance evaluation, text classification was usually used as a standard tool to determine the advantage or weakness of a text processing method, such as text representation, text feature selection, etc. In this paper, text classification is carried out to classify the Web documents collected from XSSC Website (). The performance of support vector machine (SVM) and back propagation neural network (BPNN) is compared on this task. Specifically, binary text classification and multi-class text classification were conducted on the XSSC documents. Moreover, the classification results of both methods are combined to improve the accuracy of classification. An experiment is conducted to show that BPNN can compete with SVM in binary text classification; but for multi-class text classification, SVM performs much better. Furthermore, the classification is improved in both binary and multi-class with the combined method.

Text mining, also known as discovering knowledge from the text, which has emerged as a possible solution for the current information explosion, refers to the process of extracting non-trivial and useful patterns from unstructured text. Among the general tasks of text mining such as text clustering, summarization, etc, text classification is a subtask of intelligent information processing, which employs unsupervised learning to construct a classifier from training text by which to predict the class of unlabeled text. Because of its simplicity and objectivity in performance evaluation, text classification was usually used as a standard tool to determine the advantage or weakness of a text processing method, such as text representation, text feature selection, etc. In this paper, text classification is carried out to classify the Web documents collected from XSSC Website (). The performance of support vector machine (SVM) and back propagation neural network (BPNN) is compared on this task. Specifically, binary text classification and multi-class text classification were conducted on the XSSC documents. Moreover, the classification results of both methods are combined to improve the accuracy of classification. An experiment is conducted to show that BPNN can compete with SVM in binary text classification; but for multi-class text classification, SVM performs much better. Furthermore, the classification is improved in both binary and multi-class with the combined method.

Text based Mining is the process of analyzing a document or set of documents to understand the content and meaning of the information they contain. Text Mining enhances human's ability to Process massive quantities of information and it has high Commercial values. Text mining, sometimes alternately referred to as text data mining, roughly, process of deriving highquality information from text. High-quality information is typically derived through the devising of patterns and trends through means such as statistical pattern learning. It usually involves the process of structuring the input text deriving patterns within the structured data, and finally evaluation and interpretation of the output. 'High quality' in text mining usually refers to some combination of relevance, novelty, and interestingness. Typical text mining tasks include text categorization, text clustering, concept/entity extraction, analysis and entity relation modeling (i.e., learning relations between named entities).

Bulanık c-ortalama kümeleme, literatürde farklı alanlarda kullanılan yaygın kümeleme algoritmalarından biridir. Boyut küçültme, büyük veri kümelerini, en az bilgi kaybıyla eşdeğeri olan daha küçük boyutlu veri kümelerine dönüştüren bir tekniktir. Bu makalede, boyut azalmasının farklı bulanık kümeleme teknikleri üzerindeki etkisi incelenmektedir. Bu amaçla farklı dört bulanık kümeleme algoritması kullanıldı: Bulanık C-Ortalamalar (BCO), Tip-2 Bulanık C-Ortalamalar (BCO2), Olasılıksal Bulanık C-Ortalamalar (OBCO) ve Denetimsiz Olasılıksal Bulanık C-Ortalamalar (DOBC). Boyut küçültme için verilerdeki varyansı minimum %80 açıklayan bir dizi bileşen seçildi. Boyutsallığın azaltılması için Kesik Tekil Değer Ayrıştırma (KTDA) tekniği kullanıldı. Çalışmada, ilk olarak, orijinal gerçek dünya veri kümeleri, bahsedilen dört yöntemle kümelendi. Daha sonra, bu veri kümelerinin boyutu küçültülmüş hali de yine bu dört yöntemle kümelendi. Kümeleme performansı için dört dahili kümeleme değerlendirme metriği kullanıldı. Bunlar Silhouette İndeksi (SI), Bölme Katsayısı (BK), Bölme Entropisi (BE) ve Kök Ortalama Kare Hatası (KOKH). Yöntemlerin, orijinal ve boyutu azaltılmış veri kümeleri için kümeleme performansı, karşılaştırmalı olarak sunulmaktadır. Sonuçlara göre, indirgenmiş veriler üzerinde yöntemlerin performansı, orijinal verilerden daha başarılıdır. Boyut azaltımının kümeleme başarısına katkısı en çok BCO için, en az BCO2 için elde edilmektedir.

This paper presents a new knowledge-based vector space model (VSM) for text clustering. In the new model, semantic relationships between terms (e.g., words or concepts) are included in representing text documents as a set of vectors. The idea is to calculate the dissimilarity between two documents more effectively so that text clustering results can be enhanced. In this paper, the semantic relationship between two terms is defined by the similarity of the two terms. Such similarity is used to re-weight term frequency in the VSM. We consider and study two different similarity measures for computing the semantic relationship between two terms based on two different approaches. The first approach is based on the existing ontologies like WordNet and MeSH. We define a new similarity measure that combines the edge-counting technique, the average distance and the position weighting method to compute the similarity of two terms from an ontology hierarchy. The second approach is to make use of text corpora to construct the relationships between terms and then calculate their semantic similarities. Three clustering algorithms, bisecting k-means, feature weighting k-means and a hierarchical clustering algorithm, have been used to cluster real-world text data represented in the new knowledge-based VSM. The experimental results show that the clustering performance based on the new model was much better than that based on the traditional term-based VSM.

Günümüzde insansız araçlarla alan taraması, yani bir alanın tümünün veya bir kısmının insansız araçlarla en az efor ile dolaşılması, alan taramasına duyulan ihtiyaç ve insansız araçların kullanımının artmasıyla beraber hızla önem kazanmaktadır. İnsansız araçlarla alan taramasının, İHA'lar ( İnsansız Hava Araçları) ile bir alanda keşif yapmaktan robotlar ile mayınlı arazilerin mayınlardan arındırılmasına, büyük alışveriş merkezlerinde yerlerin robotlarla temizlenmesinden büyük arazilerde çim biçmeye kadar pek çok uygulaması mevcuttur. Problemin tek araçla alan taraması, birden fazla araçla alan taraması, çevrimiçi (arazinin nasıl oldugu daha önceden bilinmiyorsa) alan taraması gibi pek çok versiyonu mevcuttur. Ayrıca arazide çeşitli büyüklüklerde engeller de bulunabilmektedir. Dogal olarak, bu problem üzerinde birçok araştırmacı çalışmaktadır ve günümüze kadar pek çok çalışma yapılmıştır. Bu çalış -maların çok büyük bir kısmı, MKA (Minimum Kapsama Agacı) yaklaşımını kullanmaktadır. Bu yaklaşımda temel olarak, düzlemsel bir arazi, aracın görüş alanına göre eş büyüklükte karelere bölünmekte ve bu karelerin merkezleri birer dügüm olarak kabul edilerek araziden kenarları birim agırlıkta olan bir çizge elde edilmektedir. Sonra bu çizgenin MKA'sı bulunup bu MKA'nın etrafı araçlar tarafından turlanmaktadır. Bizim önerdigimiz metot düzlemsel olmayan arazilerin de insansız araçlar ile taranmasına çözüm getirmektedir. Biz de çözümde MKA yaklaşımını temel aldık, ancak arazi düzlemsel olmadıgı için çizgedeki kenarlara birim agırlık vermek yerine iki kare arasındaki egime baglı agırlıklar verdik. Bu yaklaşım ile aynı zamanda özellikle İHA'lar için rüzgarın şiddeti ve yönü de hesaba katılarak bir rota elde edilip alan taraması yapılabilir.

Clustering constitutes an ubiquitous problem when dealing with huge data sets for data compression, visualization, or preprocessing. Prototype-based neural methods such as neural gas or the self-organizing map offer an intuitive and fast variant which represents data by means of typical representatives, thereby running in linear time. Recently, an extension of these methods towards relational clustering has been proposed which can handle general non-vectorial data characterized by dissimilarities only, such as alignment or general kernels. This extension, relational neural gas, is directly applicable in important domains such as bioinformatics or text clustering. However, it is quadratic in m both in memory and in time (m being the number of data points). Hence, it is infeasible for huge data sets. In this contribution we introduce an approximate patch version of relational neural gas which relies on the same cost function but it dramatically reduces time and memory requirements. It offers a single pass clustering algorithm for huge data sets, running in constant space and linear time only.

Clustering constitutes an ubiquitous problem when dealing with huge data sets for data compression, visualization, or preprocessing. Prototype-based neural methods such as neural gas or the self-organizing map offer an intuitive and fast variant which represents data by means of typical representatives, thereby running in linear time. Recently, an extension of these methods towards relational clustering has been proposed which can handle general non-vectorial data characterized by dissimilarities only, such as alignment or general kernels. This extension, relational neural gas, is directly applicable in important domains such as bioinformatics or text clustering. However, it is quadratic in m both in memory and in time (m being the number of data points). Hence, it is infeasible for huge data sets. In this contribution we introduce an approximate patch version of relational neural gas which relies on the same cost function but it dramatically reduces time and memory requirements. It offers a single pass clustering algorithm for huge data sets, running in constant space and linear time only.

In this paper, we present a particle swarm optimizer (PSO) to solve the variable weighting problem in projected clustering of high-dimensional data. Many subspace clustering algorithms fail to yield good cluster quality because they do not employ an efficient search strategy. In this paper, we are interested in soft projected clustering. We design a suitable k-means objective weighting function, in which a change of variable weights is exponentially reflected. We also transform the original constrained variable weighting problem into a problem with bound constraints, using a normalized representation of variable weights, and we utilize a particle swarm optimizer to minimize the objective function in order to search for global optima to the variable weighting problem in clustering. Our experimental results on both synthetic and real data show that the proposed algorithm greatly improves cluster quality. In addition, the results of the new algorithm are much less dependent on the initial cluster centroids. In an application to text clustering, we show that the algorithm can be easily adapted to other similarity measures, such as the extended Jaccard coefficient for text data, and can be very effective.

Information is commonly reflected in news articles. However, texts are unstructured and thus demanding to analyze automatically. To identify and capture the facts in a news story we propose a novel approach, which utilizes natural language engineering. A combination of selected linguistic and statistical criteria enables the identification of grammatical units such as noun, verb, adjective, and adverb phrases. In literature, these entities are presumed to carry the meaning and the information expressed in English texts. In our study, we focus on determining multi-word features in articles related to the monetary policy conducted by the central bank in the USA, FED. The features are composed as attribute-value pairs, where the attributes represent grammatical units, which quantify the major event characteristics. The corresponding values are conditional expressions, which vary over time as facts evolve. The final set is aggregated over the corpus by the application of heuristic and syntax-based rules. Financial experts contributed to the project by providing expertise for the document interpretation.

Web Person Disambiguation (WPD) is often done through clustering of web documents to identify the different namesakes for a given name. This paper presents a clustering algorithm using key phrases as the basic feature. However, key phrases are used in two different forms to represent the document as well context information surround the name mentions in a document. In using the vector space model, key phrases extracted from the documents are used as document representation. Context information of name mentions is represented by skip bigrams of the key phrase sequences surrounding the name mentions. The two components are then aggregated into the vector space model for clustering Experiments on the W eP S2 datasets show that the proposed approach achieved comparable results with the top 1 system. It indicates that key phrases can be a very effective feature for WPD both at the document level and at the sentential level near the name mentions.

Text Document are tremendously increasing in the internet, the hierarchical document clustering has proven to be useful in grouping similar document for large applications. Still most documents suffer from problems of high dimensionality, scalability, accuracy and meaningful cluster labels. In this paper an new approach fuzzy frequent itemsets based hierarchical clustering is proposed, in which fuzzy association rule mining algorithm is used to improve the clustering accuracy. In this approach firstly the key terms are extracted from the document set and each document is preprocessed into the document representation for the further mining process. Secondly, a fuzzy association rule mining algorithm for text is discover to find the sets, highly related fuzzy frequent itemsets, in which the key terms are regarded as a labels of the candidate clusters. Referring the candidate cluster it has been experimentally evaluated based on classic 30, Classic 4, and tr11 data sets for two methods FIHC and K-means in MATLAB 2009Rb.

The content of a text is mainly defined by keywords and named entities occurring in it. In particular for news articles, named entities are usually important to define their semantics. However, named entities have ontological features, namely, their aliases, types, and identifiers, which are hidden from their textual appearance. In this paper, we explore weighted combinations of those latent named entity features with keywords for text clustering. To that end, the traditional Vector Space Model is adapted with multiple vectors defined over spaces of entity names, types, name-type pairs, identifiers, and keywords. Clustering quality is evaluated by both of the self purity-separation type and the relative comparison type of measures. Hard and fuzzy clustering experiments of the proposed model on selected data subsets of Reuters-21578 are conducted and evaluated.

This paper adapts a widespread formalism of Knowledge Representation known in the AI literature as J. Sowa's Conceptual Graphs to the purposes of Content Analysis. It is proposed that instead of nested contexts, negation and modalities could be represented by colouring the links and the nodes of the graphs. This colour-based representation of logic enables the replacement of Sowa's game-theoretical semantics by a matching algorithm. It is also claimed that the resulting 'local' application of negation and modalities is close to the behaviour of negation and adverbs in natural language. Above all, the use of colours restores a unique level in semantic networks. These 'flat' semantic networks are appealing to social scientists because they avoid information redundancy and improve connectivity in the network. Two methods of textual analysis that can take direct benefit from semantic networks (semantic paths distributions and texts clustering) are presented. However, 'flat' semantic networks that remain semantically correct are perhaps even more interesting because they may bridge standard networks with connectionnist networks. In such a perspective, Coloured Conceptual Graphs represent an intermediate but promising step.

Obtaining correct and relevant information at the right time to user's query is quite a difficult task. This becomes even complex, if the query terms have many meanings and occur in different varieties of domain. This paper presents a fuzzy-ontology based information retrieval system that determine the semantic equivalence between terms in a query and terms in a document by relating the synonyms of query terms with those of document terms. Hence, documents could be retrieved based on the meaning of query terms. The challenge has been that surface form does not sufficiently retrieve relevant document to user's query. However, the results presented showed that the Fuzzy-Ontology Information Retrieval system successfully retrieve relevant documents to user's query. This is irrespective of different meaning and varieties of domain. The System was tested on words with different meanings and some set of user's query from varied domains.

Generally, textual data sets are represented by using different models. But, sometimes it does not capture the text structure, or some models that preserves text structure. Vector space model is also known as the ‘bag of word model’. To represent textual document using vector space model is based on most text mining methods. This model cannot maintain the text structure as it is. Compression distances are the most widely used technique for the purpose of knowledge discovery and mining of data also to improve the performance metric. Compression distance technique is for measuring the similarity between two documents. By applying Distortion technique which is for purpose of destroys the text structure. A distortion technique removes non-relevant words as well as maintains the text structure. Mostly, clustering is based on compression. The NCD i.e. Normalized compression distance technique is works in any application area. By length of compressed data files NCD gets computed. NCD captu...

Generally, textual data sets are represented by using different models. But, sometimes it does not capture the text arrangement as it is. Vector space model is also recognized as the bag of word model. To represent textual document using vector space model is based on most text mining methods. This model cannot maintain the text structure as it is. Compression distances are the most widely used technique for the purpose of knowledge discovery and mining of data also to improve the performance metric. Compression distance technique is for measuring the similarity between two documents. By applying Distortion technique which is for purpose of destroys the text structure. A distortion technique removes nonrelevant words as well as maintains the text structure. The NCD i.e. normalized compression distance technique gives the structural similarity between text documents. By applying a detailed word removal method clustering accuracy can be enhanced. In proposed system, JPG encoding is th...

Generally, textual data sets are represented by using different models. But, sometimes it does not capture the text arrangement as it is. Vector space model is also recognized as the bag of word model. To represent textual document using vector space model is based on most text mining methods. This model cannot maintain the text structure as it is. encoding technique wherein an image pixel contains the values of RGB. So, it greatly reduces the size of the document. Each pixel value assigns an equivalent English letter of the same byte value where each pixel contains 3 letters. To reduce the size of text documents, For getting the lossless compression, Encodes the text document in term of image format.

Günümüzde, karmaşık radar hedef eko sinyal ortamlarındaki gerçek hedef sayısının tespiti işlemi savunma sistemleri ve haberleşme alanında önem arz eden konular arasında yer almaktadır. Özellikle, birden fazla hedef sinyalinin bulunduğu bu karmaşık hedef alanında hedef sayılarının doğru bir şekilde belirlenmesi büyük ölçüde önem kazanmıştır. Bu çalışmada, herbirinin içinde farklı sayılarda hedef eko sinyali olan gerçek çoklu hedef eko sinyal ortamları kullanılmıştır. Kullanılan gerçek çoklu hedef eko sinyalleri, radar deney setinden elde edilmiştir. Daha sonra ise Davies-Bouldin (DB), The Xie-Beni (XB), The PBM (Pakhira, Bandyopadhyay and Maulik), and The Calinski Harabasz (CH) Bulanık C Ortalamalar (BCO) indeksleme yöntemlerinin çoklu eko sinyal bulunan ortamlardaki doğru eko sinyal küme sayısını bulma performansları birbirleri ile karşılaştırılarak değerlendirilmiştir.

Bu çalışma, Transformer mimarisinin giriş katmanlarının matematiksel temellerine odaklanmaktadır. Giriş katmanları, ham veriyi modelin işleyebilecegi anlamlı bir forma dönüştürmek için kritik bir rol oynar. Bu baglamda, tokenizasyon, embedding ve positional encoding süreçleri detaylışekilde incelenmiştir. Tokenizasyon aşamasında, ham veri belirli kurallara göre anlamlı birimlere ayrılmış, ardından bu birimler sayısal bir forma dönüştürülerek model için uygun bir giriş dizisi oluşturulmuştur. Embedding aşamasında, bu sayısal temsiller yüksek boyutlu vektörler halinde ifade edilmiş ve her bir token için anlam taşıyan vektör temsilleri elde edilmiştir. Positional encoding süreci ise sinüs ve kosinüs tabanlı fonksiyonlarla token sırasını koruyan ek bilgiler saglamıştır. Bu çalışmada, tüm bu süreçlerin matematiksel modelleri ayrıntılı olarak açıklanmış ve giriş katmanlarının Transformer'ın genel performansına olan katkısı vurgulanmıştır. Giriş katmanlarının matematiksel temellerini anlamak, Transformer mimarisinin baglam modelleme kapasitesini daha derinlemesine kavramak için bir rehber sunmaktadır.

Bu çalışma, Transformer mimarisinin giriş katmanlarının matematiksel temellerine odaklanmaktadır. Giriş katmanları, ham veriyi modelin işleyebilecegi anlamlı bir forma dönüştürmek için kritik bir rol oynar. Bu baglamda, tokenizasyon, embedding ve positional encoding süreçleri detaylışekilde incelenmiştir. Tokenizasyon aşamasında, ham veri belirli kurallara göre anlamlı birimlere ayrılmış, ardından bu birimler sayısal bir forma dönüştürülerek model için uygun bir giriş dizisi oluşturulmuştur. Embedding aşamasında, bu sayısal temsiller yüksek boyutlu vektörler halinde ifade edilmiş ve her bir token için anlam taşıyan vektör temsilleri elde edilmiştir. Positional encoding süreci ise sinüs ve kosinüs tabanlı fonksiyonlarla token sırasını koruyan ek bilgiler saglamıştır. Bu çalışmada, tüm bu süreçlerin matematiksel modelleri ayrıntılı olarak açıklanmış ve giriş katmanlarının Transformer'ın genel performansına olan katkısı vurgulanmıştır. Giriş katmanlarının matematiksel temellerini anlamak, Transformer mimarisinin baglam modelleme kapasitesini daha derinlemesine kavramak için bir rehber sunmaktadır. Here's the English translation of the summary: Abstract This study focuses on the mathematical foundations of the input layers in the Transformer architecture. Input layers play a critical role in transforming raw data into a meaningful format that can be processed by the model. In this context, the processes of tokenization, embedding, and positional encoding are analyzed in detail. During tokenization, raw data is divided into meaningful units based on specific rules, and these units are then converted into numerical representations to form a suitable input sequence for the model. In the embedding stage, these numerical representations are expressed as high-dimensional vectors, providing meaningful vector representations for each token. The positional encoding process incorporates additional information to preserve token order using sine and cosine-based functions. This study elaborates on the mathematical models of all these processes and highlights the contribution of input layers to the overall performance of the Transformer. Understanding the mathematical foundations of input layers offers a deeper insight into the contextual modeling capacity of the Transformer architecture.

We present in this work a new algorithm for document hierarchical clustering and automatic generation of portals sites. This model is inspired from the self-assembling behavior observed in real ants where ants progressively get attached to an existing support and successively to other attached ants. The artificial ants that we have defined will similarly build a tree. Each ant represents one document. The way ants move and build this tree depends on the similarity between the documents. We have tested our model on sets of web pages extracted from internet and we have successfully compared our results to those obtained by the AHC (Ascending Hierarchical Clustering).

Traditional approaches to document classification requires labeled data in order to construct reliable and accurate classifiers. Unfortunately, labeled data are seldom available, and often too expensive to obtain, especially for large domains and fast evolving scenarios. Given a learning task for which training data are not available, abundant labeled data may exist for a different but related domain. One would like to use the related labeled data as auxiliary information to accomplish the classification task in the target domain. Recently, the paradigm of transfer learning has been introduced to enable effective learning strategies when auxiliary data obey a different probability distribution. A co-clustering based classification algorithm has been previously proposed to tackle cross-domain text classification. In this work, we extend the idea underlying this approach by making the latent semantic relationship between the two domains explicit. This goal is achieved with the use of Wikipedia. As a result, the pathway that allows to propagate labels between the two domains not only captures common words, but also semantic concepts based on the content of documents. We empirically demonstrate the efficacy of our semantic-based approach to cross-domain classification using a variety of real data.

Abstract—Traditional approaches to document classification requires labeled data in order to construct reliable and accurate classifiers. Unfortunately, labeled data are seldom available, and often too expensive to obtain, especially for large domains and fast evolving ...

Abstract. In this paper we describe our participation in the Third Web People Search (WePS3) evaluation campaign. We took part in the Online Reputation Management (ORM) task. Ambiguity of organization names (eg,“Amazon” or “Apple”) raises obvious difficulties for systems that attempt to trace mentions of and opinions about a specific company in Web data, in an unsupervised manner. Problems are further amplified in the context of user generated content, where proper capitalization of named entities is often absent. The ...

In this paper we describe our participation in the Second Web People Search workshop (WePS2) and detail our approaches. For the clustering task, our focus was on replicating the lessons learned at WEPS1 on the data set made available as part of WEPS2 and on experimenting with a voting-based combination of clustering methods. We found that clustering methods display the same overall behavior on the WEPS1 and WESP2 data sets and that a hierarchical clustering approach delivers the best performance, even ...

Text clustering is an important method for organising the increasing volume of digital content, aiding in the structuring and discovery of hidden patterns in uncategorised data. The effectiveness of text clustering largely depends on the selection of textual embeddings and clustering algorithms. This study argues that recent advancements in large language models (LLMs) have the potential to enhance this task. The research investigates how different textual embeddings, particularly those utilised in LLMs, and various clustering algorithms influence the clustering of text datasets. A series of experiments were conducted to evaluate the impact of embeddings on clustering results, the role of dimensionality reduction through summarisation, and the adjustment of model size. The findings indicate that LLM embeddings are superior at capturing subtleties in structured language. OpenAI's GPT-3.5 Turbo model yields better results in three out of five clustering metrics across most tested datasets. Most LLM embeddings show improvements in cluster purity and provide a more informative silhouette score, reflecting a refined structural understanding of text data compared to traditional methods. Among the more lightweight models, BERT demonstrates leading performance. Additionally, it was observed that increasing model dimensionality and employing summarisation techniques do not consistently enhance clustering efficiency, suggesting that these strategies require careful consideration for practical application. These results highlight a complex balance between the need for refined text representation and computational feasibility in text clustering applications. This study extends traditional text clustering frameworks by integrating embeddings from LLMs, offering improved methodologies and suggesting new avenues for future research in various types of textual analysis.

We describe latent Dirichlet allocation (LDA), a generative probabilistic model for collections of discrete data such as text corpora. LDA is a three-level hierarchical Bayesian model, in which each item of a collection is modeled as a finite mixture over an underlying set of topics. Each topic is, in turn, modeled as an infinite mixture over an underlying set of topic probabilities. In the context of text modeling, the topic probabilities provide an explicit representation of a document. We present efficient approximate inference techniques based on variational methods and an EM algorithm for empirical Bayes parameter estimation. We report results in document modeling, text classification, and collaborative filtering, comparing to a mixture of unigrams model and the probabilistic LSI model.

Delivering effective customer service via the Internet requires attention to many aspects of knowledge management if it is to be convenient and satisfying for customers, while at the same time efficient and economical for the company or other organization. In RightNow eService Center, such management is enabled by automatically gathering meta-knowledge about the Answer documents held in the core knowledge base. A variety of AI techniques are used to facilitate the construction, maintenance, and navigation of the knowledge base. These include collaborative filtering, swarm intelligence, fuzzy logic, natural language processing, text clustering, and classification rule learning. Customers using eService Center report dramatic decreases in support costs and increases in customer satisfaction due to the ease of use provided by the "self-learning" features of the knowledge base.

The technology of automatic document summarization is maturing and may provide a solution to the information overload problem. Nowadays, document summarization plays an important role in information retrieval. With a large volume of documents, presenting ...

In this paper some results of a new text clustering methodology are presented. A prototype is an interesting document or a part of an extracted, interesting text. The given prototype is matched with the existing document database or the monitored document flow. Our claim is that the new methodology is capable of automatic content-based clustering using the information of the document. To verify this hypothesis an experiment was designed with the Bible. Four different translations, one Greek, one Latin, and two Finnish translations from years 1933/38 and 1992 were selected as test text material. Validation experiments were performed with a designed prototype version of the software application.

There is a vast amount of financial information on companies' financial performance available to investors today. While automatic analysis of financial figures is common, it has been difficult to automatically extract meaning from the textual part of financial reports. The textual part of an annual report contains richer information than the financial ratios. In this paper, we combine data mining methods for analyzing quantitative and qualitative data from financial reports, in order to see if the textual part of the report contains some indication about future financial performance. The quantitative analysis has been performed using selforganizing maps, and the qualitative analysis using prototype-matching text clustering. The analysis is performed on the quarterly reports of three leading companies in the telecommunications sector.

IJCSI International Journal of Computer Science Issues, Vol. 8, Issue 3, No. 2, May 2011 ISSN (Online): 1694-0814 www.IJCSI.org ... Ahmad Zuhdi1, Aniati Murni Arymurthy2 and Heru Suhartanto3 ... 1 Informatics Engineering Dept., Trisakti University and Post Graduate Student in ...

Bulanık c-ortalama kümeleme, literatürde farklı alanlarda kullanılan yaygın kümeleme algoritmalarından biridir. Boyut küçültme, büyük veri kümelerini, en az bilgi kaybıyla eşdeğeri olan daha küçük boyutlu veri kümelerine dönüştüren bir tekniktir. Bu makalede, boyut azalmasının farklı bulanık kümeleme teknikleri üzerindeki etkisi incelenmektedir. Bu amaçla farklı dört bulanık kümeleme algoritması kullanıldı: Bulanık C-Ortalamalar (BCO), Tip-2 Bulanık C-Ortalamalar (BCO2), Olasılıksal Bulanık C-Ortalamalar (OBCO) ve Denetimsiz Olasılıksal Bulanık C-Ortalamalar (DOBC). Boyut küçültme için verilerdeki varyansı minimum %80 açıklayan bir dizi bileşen seçildi. Boyutsallığın azaltılması için Kesik Tekil Değer Ayrıştırma (KTDA) tekniği kullanıldı. Çalışmada, ilk olarak, orijinal gerçek dünya veri kümeleri, bahsedilen dört yöntemle kümelendi. Daha sonra, bu veri kümelerinin boyutu küçültülmüş hali de yine bu dört yöntemle kümelendi. Kümeleme performansı için dört dahili kümeleme değerlendirme metriği kullanıldı. Bunlar Silhouette İndeksi (SI), Bölme Katsayısı (BK), Bölme Entropisi (BE) ve Kök Ortalama Kare Hatası (KOKH). Yöntemlerin, orijinal ve boyutu azaltılmış veri kümeleri için kümeleme performansı, karşılaştırmalı olarak sunulmaktadır. Sonuçlara göre, indirgenmiş veriler üzerinde yöntemlerin performansı, orijinal verilerden daha başarılıdır. Boyut azaltımının kümeleme başarısına katkısı en çok BCO için, en az BCO2 için elde edilmektedir.

We discuss our experiences in analyzing customer-support issues from the unstructured free-text fields of technical-support call logs. The identification of frequent issues and their accurate quantification is essential in order to track aggregate costs broken down by issue type, to appropriately target engineering resources, and to provide the best diagnosis, support and documentation for most common issues. We present a new set of techniques for doing this efficiently on an industrial scale, without requiring manual coding of calls in the call center. Our approach involves (1) a new text clustering method to identify common and emerging issues; (2) a method to rapidly train large numbers of categorizers in a practical, interactive manner; and (3) a method to accurately quantify categories, even in the face of inaccurate classifications and training sets that necessarily cannot match the class distribution of each new month's data. We present our methodology and a tool we developed and deployed that uses these methods for tracking ongoing support issues and discovering emerging issues at HP.

Bu calismada Meksika San Pedro Martir (SPM) Ulusal Astronomi Gozlemevi'nde 84 cm teleskop ile gozlenmis 20 acik yildiz kumesinin CCD UBVRI fotometrisi verileri ile gozlemsel parametreleri olan renk artiklari, metal ve agir element bolluklari, uzaklik modulleri, uzakliklari ve yaslari elde edilmistir. Metal bollugu tayini, $(U-B)$-$(B-V)$ iki renk diyagraminda Hyades anakolu uzerindeki metalce fakir parlak F tayf turunden yildizlarin dagilimlarindan belirlenen $\delta(U$--$B)$ morotesi artigi teknigiyle yapilmistir. 20 kumenin uzaklik modulu, uzaklik ve yaslari, esyas egrilerinin kumelerin bes renk indeksindeki ($B-V$, $R-I$, $V-R$, $V-I$, $B-R$) renk-kadir diyagramlari ile cakistirilmasindan elde edilmistir.

Decide Madrid is the civic technology of Madrid City Council which allows users to create and support online petitions. Despite the initial success, the platform is encountering problems with the growth of petition signing because petitions are far from the minimum number of supporting votes they must gather. Previous analyses have suggested that this problem is produced by the interface: a paginated list of petitions which applies a non-optimal ranking algorithm. For this reason, we present an interactive system for the discovery of topics and petitions. This approach leads us to reflect on the usefulness of data visualization techniques to address relevant societal challenges.

Analyzing financial performance in today's information-rich society can be a daunting task. With the evolution of the Internet, access to massive amounts of financial data, typically in the form of financial statements, is widespread. Managers and stakeholders are in need of a data-mining tool allowing them to quickly and accurately analyze this data. An emerging technique that may be suited for this application is the self-organizing map. The purpose of this study was to evaluate the performance of self-organizing maps for analyzing financial performance of international pulp and paper companies. For the study, financial data, in the form of seven financial ratios, was collected, using the Internet as the primary source of information. A total of 77 companies, and six regional averages, were included in the study. The time frame of the study was the period 1995-00. An example analysis was performed, and the results analyzed based on information contained in the annual reports. The results of the study indicate that self-organizing maps can be feasible tools for the financial analysis of large amounts of financial data.

Bu calismada Meksika San Pedro Martir (SPM) Ulusal Astronomi Gozlemevi'nde 84 cm teleskop ile gozlenmis 20 acik yildiz kumesinin CCD UBVRI fotometrisi verileri ile gozlemsel parametreleri olan renk artiklari, metal ve agir element bolluklari, uzaklik modulleri, uzakliklari ve yaslari elde edilmistir. Metal bollugu tayini, $(U-B)$-$(B-V)$ iki renk diyagraminda Hyades anakolu uzerindeki metalce fakir parlak F tayf turunden yildizlarin dagilimlarindan belirlenen $\delta(U$--$B)$ morotesi artigi teknigiyle yapilmistir. 20 kumenin uzaklik modulu, uzaklik ve yaslari, esyas egrilerinin kumelerin bes renk indeksindeki ($B-V$, $R-I$, $V-R$, $V-I$, $B-R$) renk-kadir diyagramlari ile cakistirilmasindan elde edilmistir.

Text Clustering is a text mining technique which is used to group similar documents into single cluster by using some sort of similarity measure & separating the dissimilar documents. Popular clustering algorithms available for text clustering treats document as conglomeration of words. The syntactic or semantic relations between words are not given any consideration. Many different algorithms were propagated to study and find connection among different words in a sentence by using different concepts. In this paper, a hybrid algorithm for clustering of Punjabi text document that uses semantic relations among words in a sentence for extracting phrases has been developed. Phrases extracted create a feature vector of the document which is used for finding similarity among all documents. Experimental results reveal that hybrid algorithm performs better with real time data sets.

Özetçe-Derin öğrenmenin doğal dil işleme problemlerinde kullanılmaya başlaması ile bu alandaki birçok problemin çözümünde ciddi iyileşmeler olmuştur. Kelime dizilerini etiketleme problemlerinin de derin öğrenme yöntemleri ile sıkça çalışılan konulardan birisi olduğu görülmektedir. Bu çalışmada derin öğrenme yöntemleri kullanarak Türkçe üzerinde kelime dizisi etiketleme problemi için bir mimari önererek, karakter, morfem ya da sözcük gösterimlerinin bu problem kapsamındaki etkisini inceliyoruz. Türkçe gibi sondan eklemeli dillerde kelimenin bir bütün olarak modellenmesi seyreklik problemini de beraberinde getirmektedir. Bundan ötürü kelimeyi bir bütün olarak ele almaktansa karakterleri üzerinden bir kelimeyi ifade etmek ya da morfem ve morfemin etiket bilgisini hesaba katmak kelime hakkında daha detaylı bilgi vermekte ve seyreklik problemini de ortadan kaldırmaktadır. Bu çalışmada mevcut derin öğrenme modellerini Türkçe için farklı kelime veya altkelime düzeyinde gösterimler kullanarak uyguladık. Kelimelerin morfem bilgisinin kullanılmasının Türkçede dizi etiketleme sonuçlarını iyileştirdiği gözlenmiştir.