Sentiment Analysis: Customer Survey Questions
About This Video
Sentiment Analysis using Natural Language Processing Algorithms are often used to understand overall Customer Satisfaction. Survey questions and their respective answers constitutes an important subset in the context of textual analysis. Often a holistic overview of Customers’ needs, or request could be upheld by understanding what most of the Customers are looking for. After initial data cleaning all answers are converted to corpus of words. These word corpuses are next converted to lower case letters. Stop-words (such as the conjunctions, interjections), stemming words (such as adjective, adverbs), numbers, contextual specific words etc. are all removed. Exploratory data analysis such as Words Clustering could shed some light on the frequency of the number of words that appeared in the answers. Hierarchical Clustering Algorithms can be used to understand the words or group of words that are used together based on their frequency distance in a sentence. Based on the frequency of the words each Cluster forms a tree of definite height. Combination of words used in a sentence and to interpret correlation amongst those words, these Cluster trees could be cut at a definite height either from Upper or from Lower Branch. Higher the correlation limit, higher are the associativity of these various words with one another. Using Bi-Gram and Tri-Gram most commonly using phrases could be identified. Various statistical libraries are used to compute the overall sentiments of the customers as well.
In This Video
Data Analyst Programmer, Freudenberg Group
I am working as a Data Analyst Programmer at Freudenberg Battery Power Systems. I have been actively involved with the Data Science Community and would like to contribute and learn more from this amazing platform. I have taken part in WiDS almost every year. However, this year was even special as I took part in the Datathon and it was an incredible learning experience as we explored some Machine Learning Models & Deep Learning Models such as Tensorflow. I am also one of the Co-organizer and host at the Metro Detroit Women in Machine Learning and Data Science. I have worked extensively on Supervised and Unsupervised Machine Learning Modeling techniques and would like to learn more on that area so we can apply some of the techniques at my organization to continue with Predictive, Prescriptive & Descriptive Analytics. One of my Research Paper’s Abstract on “Predicting Heart Diseases with Machine Learning” got selected at the Australia and New Zealand Statistical Conference and it also made it to the Finals of the INFORMS Chicago Chapter’s Impactful Analytics Prize. I would like to take this opportunity to learn more and at the same time give back to all the people of the Data Science Community.