Latent semantic indexing and convolutional neural network for multi-label and multi-class text classification

• Autores: Quispe, Oscar, Ocsa, Alexander, Coronado, Ricardo
• Formato: objeto de conferencia
• Fecha de Publicación: 2017
• Institución: Consejo Nacional de Ciencia Tecnología e Innovación
• Repositorio: CONCYTEC-Institucional
• Lenguaje: inglés
• OAI Identifier: oai:repositorio.concytec.gob.pe:20.500.12390/1302
• Enlace del recurso: https://hdl.handle.net/20.500.12390/1302; https://doi.org/10.1109/la-cci.2017.8285711
• Nivel de acceso: acceso abierto
• Materia: indexing; convolution; feature extraction; feedforward neural nets; https://purl.org/pe-repo/ocde/ford#5.08.02

The classification of a real text should not be necessarily treated as a binary or multi-class classification, since the text may belong to one or more labels. This type of problem is called multi-label classification. In this paper, we propose the use of latent semantic indexing to text representation, convolutional neural networks to feature extraction and a single multi layer perceptron for multi-label classification in real text data. The experiments show that the model outperforms state of the art techniques when the dataset has long documents, and we observe that the precision is poor when the size of the texts is small.