Ford Motor company captures voice of customer through multiple connect points like surveys, warranty claims, social media, and so on. Data from voice of customer is used as Customer Advocacy to influence actions in manufacturing, customer service, marketing, and product development, thus enhancing customer experience. One of the challenges in this Customer advocacy project is to categorize natural language textual data into company-standard customer concern codes for assessing sentiments. These concern codes are categorized into focused function areas in the organization aimed at improving performance and customer experience. In this paper, we discuss some approaches towards addressing this challenge. In this work, verbatim inputs from transactional systems in quality office, warranty claims, issues matrix, customer surveys, and social media content are used. Due to free format and diverse sources, these textual comments pose challenges with content, language, and abbreviations. Traditional approaches like TF-IDF and word counts cannot classify comments to concern codes, due to complexity of verbatim and relationships among words used. We adopted word2vec method of representing comments to derive learning vector representation. This method enables understanding of similar and equivalent words in vocabulary, customization with automobile domain specific words, and usages while preserving semantic and syntactic relationships in the context. Besides, word2vec implementation would come in handy to group misspelled words under similar vector representation with no external definition of dictionary. An appropriate learning vector representation would not only help in predicting sentiment of the comment more sensibly, but also in classifying each comment to a group of concern codes. Also, we plan to use translation techniques to extend the framework developed using comments in English into other languages.