This post is translated as "Automate The Customer Service Experience for Flight Resolutions Using Amazon Lex".
With empty journey in many areas around the world, the digitalization of the aviation industry is continuing to change.Air companies are working to reduce touch points at the airport.Best practices have been introduced to minimize the number of physical exchanges between employees and travelers.As a result, the change in customer service is accelerating as the airline strives to provide smooth and seamless experience.Customers to contact the customer service of the airline include reservations, inquiries about operating information, tracking baggage, and managing customers who frequently use airplanes.The airline wants to provide customer -centered experiences so that customers can select how to engage in order to ensure high customer satisfaction.
In this post, we explain how to use a solution that has already been built on Amazon LEX to automate the customer service of an airline that books airline tickets, and provide a faster, more connected omnichanel experience.increase.
Amazon LEX provides high -level learning functions for automatic audio recognition (ASR) to convert voice into text and natural language understanding (NLU) that recognizes the intentions of text.This allows you to build an application with a very attractive customer experience and a real conversation -type interaction.Amazon Lex for Travel offers a constructed solution, so you can quickly achieve conversational experiences with customers and increase customer satisfaction.The built bots are composed of intent, sample speech, slot type for air service use cases, and are integrated with the Amazon Connect inquiry flow.
Let's take a look at an example of conversation on airline tickets while taking up various components in the built solution.(The solutions provided are only English. The conversation in this post is the image of translating Japanese.)
In this sample conversation, customers want to book tickets.Agents collect detailed information on travel, such as departure cities, destinations, accompanying accompanying, and desired departure time.After that, the agent searches for available airplanes, presents the optimal options, and books a ticket.
AirlinesServicesBot には、航空券の予約、運航情報の確認、予約詳細の取得、予約した航空券の変更、キャンセルなど、一般的なカスタマーサービスアクティビティのインテントが含まれています。具体的には以下のインテントが含まれています。
Bot's definitions include complete conversations, along with prompts for managing conversations.Each bot is also integrated with the AWS Lambda function that contains code that simulates business logic.By integrating with Amazon Kendra, you can answer questions in natural languages during conversation.
Check the overall architecture of the solution (see the following figure).
This post contains a template that creates AWS CloudFormation stacks, including AWS resources and required AWS Identity and Access Management (IAM) rolls.By using these resources, you can use a solution that has been built on the Amazon Connect channel.
Before deploying a solution, you need the following prerequisites:
Deploy the solution in the following steps.
You can test the bot using sample data.If you deploy to the Amazon Connect instance, you can interact with the bot by calling the Amazon Connect phone number.You can also test the solution directly on the Amazon Lex V2 console using voice or text.When you interact with the bot, the Dynamodb table is updated based on the conversation.After trying the conversational flow built in advance, you can customize the bot, add an intent as needed, and integrate it with the backend system.
Let's take a look at some functions that provide construction solutions, such as processing incomplete date information, interpretation of numerical values, and contact center flows.
When answering questions about the date, such as "When did you cancel your reservation?", Customers may omit the year.For example, answer "June 25" or day of the week ("Monday").In the default behavior of Amazon Lex, the bot interprets the response as a future date.Similarly, the second example answer is interpreted as the date of the next week.Building solutions solve such an incomplete (or ambiguous) date in the case of reservation dates, etc., based on transaction business logic.In addition, implicit confirmation is made by informing the customer of the interpreted date.The following code shows how the incomplete date is interpreted.
def resolve_underspecified_date(flight_booking_date): flight_booking_date = datetime.strptime(flight_booking_date, '%Y-%m-%d').date() today = date.today() if flight_booking_date <= today:return flight_booking_date else:number_of_days_in_future = (flight_booking_date - today).daysif number_of_days_in_future > 7:return flight_booking_date.replace(year=flight_booking_date.year-1)else:return flight_booking_date-timedelta(days=7)The phrase to explain ("Wife and I) instead of the explicit numerical answer (" 2 ") to the questions that try to collect the numbers (such as" How many companions? ")))))).The constructed bot will process such a descriptive answer by taking her two -step approach to capture the slot value.
The setting contains two slots.One is a custom slot type for understanding the phrase and an embedded slot type for capturing numbers.Number optional slot.In the first step, bots try to solve them using phrases containing the synonyms of each number (for example, "wife and me" is mapped to "2").If the input is not interpreted in the phrase, the bot will re -induce the guide prompt ("How many companions are there? You can answer three companions").
The next screenshot shows the value of the NumberFtravellers custom slot type.
The next screenshot shows two slots to capture one number.
The following Lambda function code interprets the numbers from conversation phrases.
number_of_travellers_in_words = dialog.get_slot( 'NumberOfTravellersInWords', intent, preference='interpretedValue')number_of_travellers = dialog.get_slot( 'NumberOfTravellers', intent, preference='interpretedValue')if departure_date and not number_of_travellers_in_words: previous_slot_to_elicit = dialog.get_previous_slot_to_elicit(intent_request) if previous_slot_to_elicit == 'NumberOfTravellersInWords':prompt = prompts.get('NumberOfTravellers')return dialog.elicit_slot('NumberOfTravellers', active_contexts, session_attributes, intent,[{'contentType': 'PlainText', 'content': prompt}]) else:prompt = prompts.get('NumberOfTravellersInWords')return dialog.elicit_slot( 'NumberOfTravellersInWords', active_contexts,session_attributes, intent, [{'contentType': 'PlainText', 'content': prompt}])if number_of_travellers or number_of_travellers_in_words \and not preferred_departure_time: if not number_of_travellers:number_of_travellers = number_of_travellers_in_wordsprompt = prompts.get('PreferredDepartureTime')return dialog.elicit_slot('PreferredDepartureTime', active_contexts, session_attributes, intent,[{'contentType': 'PlainText', 'content': prompt}])You can deploy pre -built solutions as part of the Amazon Connect inquiry flow.If a customer calls the contact center, the inquiry flow to be processed will be the inquiry flow assigned to the phone number that the customer calls.In the inquiry flow, you call an Amazon Lex bot using a block that gets the customer input.The following figure shows the inquiry flow.
Delete all the created resources so that there will be no charge in the future.
Amazon Lex for Travel promotes the provision of seamless experiences connected to customers and provides a constructed solution to promptly respond to customer requests.In this article, we have confirmed solutions for airline customers to automate tasks, such as reservations for airline tickets, obtaining operation information, and renewing reservations.The constructed solution also provides a contact center settings that can be deployed immediately on Amazon Connect.By adding conversation flows that meet your business needs, you can easily expand solutions.By building on AWS, major travel agencies can quickly respond to changing customer needs while reducing costs.Please try the air service solution that has been built on Amazon LEX!
JAYA PRAKASH KOMMU is SmartBots.AI team technology leader.We manage the team as an AI engineer who builds next -generation conversational AI interface.I enjoy soccer when I haven't designed a bot.
Sandeep Srinivasan is a product manager of the Amazon LEX team.As an eager observer in human behavior, he is passionate about customer experience.He spends his time at the time of people, technology, and the future intersection.
The translation was in charge of Ming Yang of Solution Architect.The original text is here.