Search by:
The customer sentiment analysis method from natural language texts
Full text (PDF)
UDC: 004.37
Publication Language: Ukrainian
Stuc. intelekt. 2018; 23(3): 18-26
Abstract: The article is devoted to the method of analysis of texts in the natural language, containing reviews of clients. The method differs from the existing combination of different types of vectorizer and the introduction of the component hierarchy. The sequencing of the use of different vectorizers allows us to build a hierarchy of features and markers. Using the reference vectors and island clustering techniques, with the subsequent training of a model for prediction of feelings, is one of the best methods for analyzing mood, both for non-binary and binary aspects. Based on open data set with Python and Tablau, a software product was developed to analyze customer preferences and visualize the results of analyzes.
Keywords:
References:
- Pang, B., & Lee, L. (2008). Opinion mining andsentiment analysis. Foundations and Trends® inInformation Retrieval, 2(1–2), 1-135.
- Liu, B., & Zhang, L. (2012). A survey of opinionmining and sentiment analysis. In Mining textdata (pp. 415-463). Springer, Boston, MA
- Danyliuk, I.H. Tekhnolohiia avtomatychnohovyznachennia tematyky tekstu (2008) [Tekst].Linhvistychni studii: Zb. nauk. prats. Vypusk 17 / Ukl.: Anatolii Zahnitko (nauk. red.) ta in. –Donetsk : DonNU. – S. 290-293
- Kolcz, A. Improved Robustness of SignatureBased Near-Replica Detection via LexiconRandomization (2004) / A. Kolcz, A. Chowdhury,J. Alspector. // KDD.
- Broder, Andrei Z. Identifying and Filtering NearDuplicate Documents (2000). Proceedings of the11th Annual Symposium on CombinatorialPattern Matching (COM’00). – Р. 1-10.
- J. Pennington R. Socher Chr. D. Manning. GloVe:Global Vectors for Word Representation.Available from: https://nlp.stanford.edu/pubs/glove.pdf