How sentiment analysis helps airlines better analyze data from multiple sources to enhance customer experience
By Vivek Gupta
The exponential growth of unstructured data in recent years has led businesses to embrace a data-driven approach for generating actionable insights. Industries like travel and hospitality depend heavily on data from user reviews to identify problem areas, measure success, and make strategic decisions. Understanding the sentiments expressed in these user reviews is the first step towards improving the customer experience. Here’s where sentiment analysis, powered by Natural Language Processing (NLP), can add tremendous value.
Sentiment analysis automatically analyzes product/service reviews and identifies the sentiment or opinion expressed in them. In cases where the input data is text, the sentiments identified can be positive, negative, or neutral. If the input data is voice, sentiments can vary across a wide range of emotions — happy, sad, angry, frustrated, etc. Therefore, the techniques used to identify sentiments depend on the type of data available and require appropriate modeling approaches to be developed. This blog outlines how a Fortune 500 American airline transitioned from manual processing of user reviews to an automated sentiment analysis architecture to understand user concerns and improve decision-making.
Addressing business needs through aspect-based sentiment analysis
The airline captures large volumes of user reviews covering multiple areas like staff service, food and beverages, seat comfort, inflight entertainment, etc. However, they were finding it difficult to manage growing volumes of review data from multiple sources (like social media and web forums) and correlate the analysis of these with external datasets such as weather. They wanted to automate the process of analyzing and resolving each customer review.
The Impetus team developed an aspect-based sentiment analysis solution to analyze the text from customer reviews and extract meaningful information like topics of interest (aspects), and context (sentiments).
The solution’s capabilities include:
· Analyzing heterogeneous sources of textual data
· Integrating customer data with external datasets
· Extracting aspects from a consolidated repository and filtering out irrelevant aspects
· Calculating polarity/sentiment(s) from customer reviews
Extracting actionable insights from multiple datasets
The solution’s input datasets included user reviews collected by crew members, email feedback, survey data collected by customer care teams, and tweets. These reviews contained multiple sentences, which were further broken down to capture all aspects and the associated sentiments. Next, a consolidated repository of all user reviews was created, and the following steps were performed:
· Aspect extraction:
- Collated data from all sources to create a 360-degree view
- Preprocessed data to filter out stop words/non-ASCII characters
- Applied Stanford parser to extract noun phrases
- Applied Market Basket Analysis to extract frequently appearing aspect pairs
- Retained aspects whose frequency was higher than the supported threshold and discarded the rest
· Opinion mining/sentiment analysis:
- Applied Stanford governor-dependency parser to generate governor-dependent aspect pairs for each review
- Assigned the polar word score to the aspect in cases where the governor was an aspect, and the dependent was a polar word (or vice versa)
- Reversed the polarity/sentiment of the aspect in cases where the polar word was dependent on a negative word
- Assigned the polarity/sentiment of the aspects to user reviews
Driving strategic business benefits
The NLP-powered sentiment analysis solution enabled the airline to realize several business benefits. They were able to effortlessly classify customer sentiments and identify aspects route-wise. Additionally, they could monitor and respond to unstructured consumer feedback in a systematic and timely manner. This, in turn, helped the airline better understand customer opinion and effectively improve specific aspects of the customer experience.
This is just one instance of how sentiment analysis can help enterprises effectively gain deeper business insights, resolve customer challenges, avoid customer churn, and boost growth and profitability. Impetus Technologies has helped several Fortune 1000 enterprises develop advanced analytics solutions and machine learning models for massive volumes of data. Our proven techniques can help you turn petabytes of raw data into actionable insights to drive powerful business outcomes. To learn more, get in touch today.
