An Intelligent and Scalable Search Tool For Aviation Industry

Kaushik Shakkari
Cognistx’s SQUARE/DIP Blog
5 min readSep 20, 2022


Photo by Pascal Meier on Unsplash

Check out my Open Domain Question Answering series for more context and technical details reading question answering modeling.

Aviation companies have been collecting a lot of unstructured data, like documents related to historical defects, aircraft manuals, safety reports, and troubleshooting approaches. Managing this unstructured data is essential as companies highly rely on them for different use cases. In this article, I will explain how an intelligent system can search information across various documents, which is necessary for multiple stakeholders in the aviation industry.

Use Case 1: Search Tool for Technical Service Engineers

Photo by ThisisEngineering RAEng on Unsplash

Every day technical service engineers (TSEs) troubleshoot issues related to malfunctions of multiple aircraft parts. To do that, they need to search information manually across millions of historical defect documents. This information can help them to perform best practices while troubleshooting malfunctions. However, it is tedious to search for information manually at scale.

Automating the search for troubleshooting defects can save 14,600 hours for 80 technical service engineers in a year— Singapore Airlines

Typically, documents contain a lot of technical jargon and acronyms. A simple search tool that matches exact terms is not efficient. For example, a service engineer might use a simple pdf viewer with a keyword match feature to search “Inertial Navigation System” in documents and find no results. However, in the aviation industry, INS is an acronym for Inertial Navigation System, and documents might use the INS keyword instead of Inertial Navigation System. Hence, an intelligent system that understands user query intent and domain is required to accurately automate manual searches to save time and money for airlines significantly.

Screenshot from author: Results from Cognistx’s SQUARE prototype for a user query - Did Etihad suffer from No Fault Found in 2017?

The above screenshot is from Cognistx’s SQUARE for a user query regarding “No Fault Found.” Even though the document for the second answer (highlighted in red square in the above screenshot) mentions “No Fault Found” as “NFF rate,” the model understood NFF is an acronym for No Fault Found and predicted the relevant answer.

Use case 2: Search Tool for Pilots

Photo by Blake Guidry on Unsplash

Airbus AI published multiple papers regarding how an intelligent search system can help pilots to find information in the “flight crew aircraft operations (FCOM)” documents which are typically several PDFs of several thousand pages for each aircraft.

Even experienced pilots who are very familiar with documents have difficulty finding information in constrained time — Airbus AI

Screenshot from author: High-level architecture of Cognistx’s SQUARE

Fine-tuning to enhance performance: (technical — feel free to skip)
Aeronautical documents are complex with domain-specific vocabulary, and pre-trained models trained on generic data won't give good accuracy for these documents. The paper states multitask learning boosted the performance of the model.

We have used many fine-tuning approaches for SQUARE, which include

  1. Multitask Learning
  2. Fine-tuning on downstream tasks (which require labels from domain experts or by crowdsourcing quickly with our Feedback Analytics System — FEAST)
  3. Unsupervised tuning techniques like Generative Psuedo Labeling (which requires no labels, training data is automatically generated to tune models)
  4. Re-ranking (training a model to re-rank answers based on their relevance to user query).

An ensemble of tuning techniques helped us increase performance for most use cases.

Are you interested in learning more about the technical details of fine-tuning? Feel free to reach me on LinkedIn.

Papers further stated that an ideal search system could be a speech-to-text conversational system that can take queries in natural language and give accurate responses, as voice is a more accessible medium for asking a question than typing.

Use Case 3: Search Tool for Customer Service

Airline customer service is getting worse, according to a new survey of passengers — Forbes

Image by Slash RTC from Pixabay

Customer service plays a crucial role in selecting an airline for travel. Traditionally, airline prices used to be a vital factor in choosing an airline. Interestingly, a survey conducted in 2018 stated that the top three factors for airline service are confirmed seats on the flight, arrival time, and customer service making customer onboarding and service a critical factor for customer satisfaction.

Source: Forbes

It is highly likely that if customers cannot talk to customer service or don’t get any response to their issue in time, they will avoid choosing the airline for their next trips. Both accuracy and speed of response are critical to resolving the customer's concern. More than email or call, which might take longer to connect, AI-driven chatbots can be a potential method where customers can easily interact and resolve their issues. Moreover, chatbots can understand the severity and type of the problem and direct it to the right customer care representative for further assistance.


I listed the top 3 use cases from initial research where an intelligent search tool can help the aviation industry. SQUARE could solve the above-listed pain points with high accuracy, low latency, and less cost at scale.

If you are interested in our product: CONTACT US TODAY FOR MORE DETAILS OR A DEMO!

More about Cognistx’s SQUARE (Scalable Question Answering and Recommendations Engine)

SQUARE is our scalable and production-ready intelligence search system that takes user queries and provides granular results across millions of documents in just a few seconds.

Our Aerospace Case Study:

Partnering with SAE, we have digitized all automotive and aerospace standards to help engineers locate them quickly. SQUARE extracts relevant information to allow users to find specification documents instantly. Working with prominent aerospace companies like Boeing, Pratt & Whitney, and NASA, SQUARE drives OnQue™, SAE’s next-generation search system.

Analytical Objectives:

  • Digitization of Industry Standards
  • Data Extraction from Unstructured Sources
  • AI-Driven Search and Retrieval

Add me on LinkedIn. Happy Learning!



Kaushik Shakkari
Cognistx’s SQUARE/DIP Blog

Senior Data Scientist | My biggest success till day was converting my passion to my profession (Data Science 🚀)