This study was aimed to calculate valence of customer sentiment and then mapping the sentiments into application attributes to reveal attribute-sentiment relationships. The approach used in this study illustrated the use of text mining methods to get insights from review data, which was valuable to generate recommendations for mobile application development. The sentiment analysis used lexicon-based polarity term combined with negator and amplifier. The polarity term in sentiment analysis were then mapped into identified application attributes using dependency parsing combined with lemmatization, pos tagging, and tokenization. We applied the proposed method on customer reviews of an airline mobile application (Garuda Indonesia Mobile App) scraped from Google Play store and Apple App Store. The result showed that the valence of sentiment from customer reviews have a positive relationship with star rating and negative relationship with the number of reviews. This study also indicated a number of application attributes considered relevant by users and their valence (e.g. membership, reservation and speed were the airline app's top three attributes with negative valence).