Approach for Multi-Label Text Data Class Verification and Adjustment Based on Self-Organizing Map and Latent Semantic Analysis

Abstract

In this paper, a new approach has been proposed for multi-label text data class verification and adjustment. The approach helps to make semi-automated revisions of class assignments to improve the quality of the data. The data quality significantly influences the accuracy of the created models, for example, in classification tasks. It can also be useful for other data analysis tasks. The proposed approach is based on the combination of the usage of the text similarity measure and two methods: latent semantic analysis and self-organizing map. First, the text data must be pre-processed by selecting various filters to clean the data from unnecessary and irrelevant information. Latent semantic analysis has been selected to reduce the vectors dimensionality of the obtained vectors that correspond to each text from the analysed data. The cosine similarity distance has been used to determine which of the multi-label text data class should be changed or adjusted. The self-organizing map has been selected as the key method to detect similarity between text data and make decisions for a new class assignment. The experimental investigation has been performed using the newly collected multi-label text data. Financial news data in the Lithuanian language have been collected from four public websites and classified by experts into ten classes manually. Various parameters of the methods have been analysed, and the influence on the final results has been estimated. The final results are validated by experts. The research proved that the proposed approach could be helpful to verify and adjust multi-label text data classes. 82% of the correct assignments are obtained when the data dimensionality is reduced to 40 using the latent semantic analysis, and the self-organizing map size is reduced from 40 to 5 by step 5.

A Combined Approach for Multi-Label Text Data Classification

“…In the first step, the analyzed multilabel text data must be adjusted according to the approach proposed by Kurasova and Stefanovič [ 12 ]. In this way, the quality of the data could be improved.…”

“…It is hard to decide which two classes represent text the best; it is subjective and usually depends on the opinion of the expert who assigned the class manually. For this reason, in our experimental investigation, we used the original multilabel text data and compared the results with the text data, where the class was adjusted using the approach proposed by Stefanovič and Kurasova [ 12 ]. The proposed approach allows automatic multilabel text data class adjustment using latent semantic analysis [ 23 ] and self-organizing map [ 24 ].…”

“…The proposed approach allows automatic multilabel text data class adjustment using latent semantic analysis [ 23 ] and self-organizing map [ 24 ]. The optimal threshold value was selected to be 85%, as recommended by previous research results [ 12 ].…”

“…The experimental investigation showed that the use of the multilabel text data adjustment approach suggested in the research [ 12 ] helps to adjust the original text data class and allows the classifiers to increase the accuracy. The smallest mismatch (10.90%) and the highest match (89.10%) are obtained using naive Bayes, and the perfect match is equal to 21.67%.…”

“…In the case of automatic extraction, the researcher does not have to care about the context of each class, which makes the proposed approach independent of the subjectivity of the person. Furthermore, this research used the multilabel text data adjustment model proposed by Kurasova and Stefanovič [ 12 ], which allows improving the quality of the data, as a result, it helps to achieve a higher accuracy of the classification of multilabel text data. The results of the experimental investigation have shown that the combination of classification and similarity measures can be useful and could help other researchers solve such types of classification problem.…”