Wyniki wyszukiwania - BazTech

1

Bag of words and embedding text representation methods for medical article classification

Cichosz Paweł

International Journal of Applied Mathematics and Computer Science

|

2023

|

Vol. 33, no. 4

603--621

EN

Text classification has become a standard component of automated systematic literature review (SLR) solutions, where articles are classified as relevant or irrelevant to a particular literature study topic. Conventional machine learning algorithms for tabular data which can learn quickly from not necessarily large and usually imbalanced data with low computational demands are well suited to this application, but they require that the text data be transformed to a vector representation. This work investigates the utility of different types of text representations for this purpose. Experiments are presented using the bag of words representation and selected representations based on word or text embeddings: word2vec, doc2vec, GloVe, fastText, Flair, and BioBERT. Four classification algorithms are used with these representations: a naive Bayes classifier, logistic regression, support vector machines, and random forest. They are applied to datasets consisting of scientific article abstracts from systematic literature review studies in the medical domain and compared with the pre-trained BioBERT model fine-tuned for classification. The obtained results confirm that the choice of text representation is essential for successful text classification. It turns out that, while the standard bag of words representation is hard to beat, fastText word embeddings make it possible to achieve roughly the same level of classification quality with the added benefit of much lower dimensionality and capability of handling out-of-vocabulary words. More refined embeddings methods based on deep neural networks, while much more demanding computationally, do not appear to offer substantial advantages for the classification task. The fine-tuned BioBERT classification model performs on par with conventional algorithms when they are coupled with their best text representation methods.

2

A case study in text mining of discussion forum posts: Classification with bag of words and global vectors

Cichosz P.

International Journal of Applied Mathematics and Computer Science

|

2018

|

Vol. 28, no. 4

787--801

EN

Despite the rapid growth of other types of social media, Internet discussion forums remain a highly popular communication channel and a useful source of text data for analyzing user interests and sentiments. Being suited to richer, deeper, and longer discussions than microblogging services, they particularly well reflect topics of long-term, persisting involvement and areas of specialized knowledge or experience. Discovering and characterizing such topics and areas by text mining algorithms is therefore an interesting and useful research direction. This work presents a case study in which selected classification algorithms are applied to posts from a Polish discussion forum devoted to psychoactive substances received from home-grown plants, such as hashish or marijuana. The utility of two different vector text representations is examined: the simple bag of words representation and the more refined embedded global vectors one. While the former is found to work well for the multinomial naive Bayes algorithm, the latter turns out more useful for other classification algorithms: logistic regression, SVMs, and random forests. The obtained results suggest that post-classification can be applied for measuring publication intensity of particular topics and, in the case of forums related to psychoactive substances, for monitoring the risk of drug-related crime.

3

Classification of text documents by using expanded terms in Latent Semantic Analysis

Śmiałkowska B., Gibert M.

Theoretical and Applied Informatics

|

2013

|

Vol. 25, No. 3-4

239--250

EN

In this article attention is paid to improving the quality of text document classification. The common techniques of analysis of text documents used in classification are shown and the weakness of these methods arc stressed. Discussed here is the integration of quantitative and qualitative methods, which is increasing the quality of classification. In the proposed approach the expanded terms, obtained by using information patterns are used in the Latent Semantic Analysis. Finally empirical research is presented and based upon the quality measures of the text document classification, the effectiveness of the proposed approach is proved.

PL

W artykule skoncentrowano się na poprawie jakości klasyfikacji dokumentów tekstowych. Zostały przybliżone najpopularniejsze techniki analizy dokumentów tekstowych wykorzystywanych w klasyfikacji. Zwrócono uwagę na słabe strony opisanych technik. Omówiono możliwość integracji metod ilościowych i jakościowych analizy tekstu i jej wpływ na poprawę jakości klasyfikacji. Zaproponowano rozwiązanie, w którym rozbudowane wyrażenia otrzymane za pomocą wzorców informacyjnych są wykorzystywane w niejawnej analizie semantycznej. Ostatecznie w oparciu o miary jakości klasyfikacji dokumentów tekstowych zaprezentowano wyniki badań testowych, które potwierdzają skuteczność zaproponowanego rozwiązania.

4

Propozycja mieszanego przetwarzania półstrukturalnego modelu opisu zdarzeń z akcji ratowniczo-gaśniczych Państwowej Straży Pożarnej PSP3

Mirończuk M., Maciak T.

Bezpieczeństwo i Technika Pożarnicza

|

2013

|

Nr 1

95-106

PL

W opracowaniu przedstawiono aktualnie rozwijane reprezentacje wiedzy i sposoby opisów zdarzeń, dla systemu wnioskowania na podstawie przypadków zdarzeń służb ratowniczych Państwowej Straży Pożarnej PSP. W artykule zaproponowano sposób ich przetwarzania. Przedstawiony sposób bazuje na klasyfikacji i wyszukiwaniu opisów zdarzeń.

EN

This paper describes a review of actual developed knowledge representation and case representation for fire services cases based reasoning system. The article also describes a method of processing the cases of events. This processing method based on classification and information retrieval.

5

The Concept Of Topological Information In Text Representation

Susmaga R., Masłowska I., Budzyńska L.

Foundations of Computing and Decision Sciences

|

2011

|

Vol. 36, No. 1

57-78

EN

This paper studies the possibility of processing text documents using topological information on keywords, by which we mean internal positions of the keywords in the text. While the word counts are pieces of information that is independent of the sequence of words in the text, the topological, i.e. position-related, information manifests obvious dependency on the sequence of words. In result, the presented method stops treating the texts as amorphous collections of words and starts treating them as linearly-ordered sequences of words. Thus, the introduced, topological approach is of higher level than the popular bag-of-words approaches, and its advantage should unveil in applications to texts of similar themes; due to their similar counts of keywords the topological information may prove to be indispensable. It should also require significantly smaller sets of keywords as compared to the bag-of-words approaches.

6

Text classification using word sequences

Chudzian P.

Studia Informatica : systems and information technology

|

2008

|

Vol. 1(10)

75--85

EN

The article discusses the use of word sequences in text classification. As opposed to ngrams, word sequences are not of a fixed length and therefore allow the classifier to obtain flexibility necessary to operate on documents collected from various sources. Presented classifier is built upon the suffix tree structure which enables word sequences to take part in classification process. During classification, both single words and longer sequences are taken into account and have impact on the category assignment with respect to their frequency and length. The Suffix Tree Classifier and well known Naive Bayes Classifier are compared and their properties are discussed. Obtained results show that incorporating word sequences into text classification can increase accuracy and reveal some interesting relations between maximal length of used sequences and classifier's error rate.