Recently, ACM SIGIR2018, the top conference in the field of information retrieval, was successfully held in Ann Arbor, Michigan, THE United States. Didi’s technical team deeply participated in the conference and hosted a symposium on intelligent transportation Informatics, which introduced Didi’s exploration and practice in the field of transportation in detail and shared industry-university-research cooperation experience. Didi also said that in the future, it will actively cooperate with researchers to conduct more extensive academic research and jointly solve world-class transportation and environmental protection challenges.
Ai technology is changing transportation on multiple levels
ACM SIGIR, now in its 41st year, is the international Computer Society’s premier conference on information retrieval. Each year SIGIR brings together top information retrieval researchers and practitioners from around the world to showcase the latest technologies and discoveries. In SIGIR 2018, a total of 736 papers were received, 184 papers were accepted (including 409 long papers submitted and 86 accepted), and nearly 800 people attended the conference.
The paper “Taxi or Hitchhiking: Predicting Passenger’s Preferred Service on Ride Sharing Platforms” written by Didi technical team was also included in the conference. By modeling users’ travel choices, this paper proposes a recommendation system based on users’ time, space and behavior characteristics, which can help solve the prediction and recommendation problems of users’ travel needs and preferences. Offline simulation shows that the model can greatly improve the accuracy and help users to make more efficient travel planning.
Didi algorithm experts on the spot to explain the paper, attracted many domestic and foreign counterparts and experts and scholars to exchange
Not only is the paper received, the conference’s keynote report also attracted attention from the outside world. At SIGIR 2018, Professor Ye Jieping, vice President of Didi and head of AI Labs, delivered a keynote speech detailing how Didi is using AI technology to help improve user travel experience and solve global transportation challenges. It also focuses on didi’s practical experience in intelligent delivery, intelligent map, intelligent customer service, voice recognition, intelligent transportation and other fields. In Ye Jieping’s opinion, the changes of artificial intelligence technology in transportation are multi-level, and there will be historic changes in transportation infrastructure, vehicle transportation and shared travel in the future. Didi has already been actively planning: This includes not only the underlying AI algorithms and core AI technologies, such as voice, natural language processing, and image, but also the applications supported by AI technologies — improving user experience, helping cities build smart transportation networks, and actively deploying intelligent driving and new energy vehicles.
Professor Ye Jieping, head of Didi AI Labs, talked about Didi’s AI layout and technological innovation at SIGIR
In the field of urban traffic, Didi has cooperated with more than 20 cities, including Jinan, Guiyang, Shenyang, Nanjing and Wuhan, and helped optimize more than 1,300 smart traffic lights, reducing congestion time by 10-20 percent on average, according to the report. “We will continue to invest in and cooperate extensively to extend ai applications to social welfare and make the technology create more value,” Ye said.
Industry-learning linkage to speed up intelligent transportation innovation
The integration of academia and industry was also a major focus of the conference. Vp of Didi & leader of intelligent travel Department qie Xiaohu shared didi’s industry-university-research cooperation experience during SIGIR Big data industry-university-research exchange conference. He said that Didi has rich data accumulation, and based on leading big data and technological advantages, Didi is continuously pushing forward the frontiers of technology, while actively working with the academic community to propose solutions for new sustainable development.
Director of Didi Intelligent travel Department qie Xiaohu explained in detail didi’s scientific research cooperation experience and talent cultivation mechanism on site
In addition to opening desensitization data resources and computing infrastructure to the academic community, Didi has also established research partnerships with more than 10 domestic and foreign academic research institutions, including the University of Michigan, Stanford University ARTIFICIAL Intelligence Laboratory, China Computer Society, Hong Kong University of Science and Technology, and Electrical and Electronic Engineering Association, the COMPANY said. We will jointly explore, exchange and cultivate talents in artificial intelligence, intelligent transportation, intelligent driving, economics and operational research, and jointly boost the transportation industry to make breakthroughs. Didi will also announce a new round of thematic research plans in the near future, hoping to cooperate with more outstanding scholars in professional fields.
Professor Pascal Van Hentenryck of the University of Michigan shared his thoughts on the future of mobility
It is worth noting that didi also hosted a symposium on intelligent transportation Informatics and called for papers to share the cutting-edge exploration in the field of urban traffic governance with Professors Pascal Van Hentenryck of the University of Michigan and Ben Gang of the University of Washington, as well as discuss the future of intelligent transportation informatics. After selection by the organizing committee, seven papers were finally presented in the seminar.
At the workshop, DiDi has written two papers, POI Semantic Model with a Deep Convolutional Structure and DiDi Ride Cancellation Smart Fault Determination System focuses on the research results and practices in the field of POI (information point) retrieval and decriminalization. Taking POI retrieval as an example, Didi technical team proposed a Deep POI Semantic Model based on Deep convolutional network, which can map the input text into Semantic vector space through the Deep network Model, so as to calculate the similarity between vectors and the correlation between texts. It can effectively solve the correlation matching problem in POI retrieval scenarios and improve the satisfaction of didi map retrieval. However, when the judgment is removed, a hybrid algorithm based on machine learning + rules is innovatively proposed, which can introduce tens of thousands of dimensional features as input to obtain higher accuracy and recall rate, and significantly improve user experience.
Shanghai Jiao Tong University Assistant Professor Zhang Weinan on-site speech, the seminar showed SIGIR community how to empower a new field of information retrieval topics, the content is very novel. Professor Ben Xuegang of The University of Washington also commented that Didi continues to promote data openness to the academic community, promote cooperation between universities and universities, and work with the academic community to identify and define problems, which is worth learning from more enterprises.