美团

Baidao helps Meituan’s artificial intelligence

Company Profile

Meituan.com [1] is fully known as “Beijing Sankuai Online Technology Co., Ltd.”. It is a group-buying website established on March 4, 2010. Meituan.com has the service promotion tenets of “All about food, drinks, fun and entertainment” and “One purchase on Meituan, one pleasant experience”. It is registered in Haidian District, Beijing.

industry

Online shopping

Case Background and Challenges

Meituan Waimai Domino is an online food ordering platform under Meituan. It was officially launched in November 2013, with its headquarters located in Beijing. The number of Meituan Waimai users has reached 250 million, the number of cooperative merchants exceeds 2 million, the number of active delivery riders exceeds 500,000, it covers more than 1,300 cities, and it completes 21 million orders per day. On June 27, 2019, Meituan Delivery, an instant delivery brand under Meituan, signed a strategic cooperation agreement with TOWER ENERGY. The two parties plan to promote in – depth cooperation in business areas such as electric vehicle battery – swapping services for food delivery and smart city construction.

The initial pain points of Meituan’s order processing robot lie in the fact that consumers’ consumption behaviors are highly variable, and their intentions can change at any time. One must quickly keep up with the changes in consumption behaviors in order to handle order issues in a timely manner. Dialogflow can expand and accommodate the required number of intents, that is, keep up with the changes in consumer behaviors through the design of conversations, enhancing the user experience and improving the efficiency of order processing.

Traditional computer interfaces require predictable and structured input to function properly, which makes these interfaces unnatural to use and sometimes difficult to operate. If end – users cannot easily understand this structured input, it will be hard for them to determine what actions to take. Ideally, the Meituan interface could infer the needs of end – users based on the natural language they use.

The following figure shows the processing flow of Meituan’s interaction using the API method:

The architecture is as follows:

Description of the solution architecture

1. The end user inputs or speaks the content of the expression.
2. The customer’s service sends this end – user expression to Dialogflow through the detect intent request message.
3. Dialogflow sends a detect intent response message to the service. This message contains information about the matched intent, action, parameters, and the response defined for that intent.
4. The request service performs actions as needed, such as database queries or external API calls.
5. The request service sends a response to the end user.
6. The end user sees or hears the response and then makes a relevant reply.

Use the product

Dialogflow

Features of the plan

Dialogflow adopts an open – ended approach, which enables Meituan to easily adopt new channels, eliminating costly integration work. At the same time, the platform is concise, intuitive, well – documented, and easy to use. Choosing Dialogflow is a natural thing to do. After that, the customer will combine Cloud Natural Language, and Dialogflow will conduct sentiment analysis on each user query. Then, based on the sentiment analysis score, dissatisfied users will be transferred to online customer service staff, or a better understanding of which intents generate the most customer emotions can be achieved.