Scania • Södertälje • Heltid
In the coming years the transport system for goods and people will undergo significant changes. Upcoming technological changes will reshape the value network and have a significant impact on future business models. Scania is undergoing a transformation from being a supplier of trucks, buses and engines to a supplier of complete and sustainable transport solutions. A critical success factor in the shift to becoming a world-leading provider of transport solutions is operational availability of vehicles. The belief is that the importance of uptime will be even further accentuated going forward.
Artificial intelligence (AI) is increasingly being adopted across industries. Some of the most promising AI applications so far are leveraging natural language processing (NLP). At Scania we want to leverage NLP techniques to better understand our customer and products. More specifically in this thesis project we want to understand why our customers perform repairs in our workshop network and how that knowledge in turn could be used for troubleshooting.
In Scania’s workshop network large amount of unstructured text data is generated every day. Each fault claim contains human-generated written data. That data potentially carries important information about what problems our customers experience and the symptoms that make them aware of the problems.
Fault claims submitted by our customers are reaching Scania's offices from all around the world. The claims are written in tens of different languages and regard hundreds of different problems. By using automated services, the claims can be translated into English, where supervised-learning-based text classifiers (e.g., Recurrent or Convolutional Neural Networks) can be trained to detect the problem. Multilingual, deep-learning methods (see https://arxiv.org/pdf/1911.02116.pdf for such a model), calibrated on only a sample of cases, are capable to keep the detection accuracy high while disregarding the translation step. Can such models be improved if we pre-train (not only calibrate) them on millions of cases? Can we improve them further with new in-domain objectives? How can we improve the accuracy of under-represented languages and infrequent classes? Can we use existing lexicon resources to yield highly interpretable models? These are some of the questions to be investigated during this thesis.
The project will develop knowledge and methods to be used in data-driven diagnostics. The NLP methods will be applied to texts retrieved from workshop work orders and should allow for effective implementation of NLP powered troubleshooting.
Education and skills
Master’s student in computer science, computational linguistics, mathematics or similar, preferably with specialization or interest in statistics, machine learning and NLP. Documented experience and skills in Python is a merit.
A thesis project is a great way to learn more about Scania and our many interesting career opportunities.
Number of students: 1-2 (you can apply separately or together)
Start date: Autumn 2021
Estimated time needed: 20 weeks (30 credits), Full-time
Language of work: Good knowledge in English is required
The work will be carried out at our offices in Södertälje and from home.
Enclose CV, personal letter, and grades. If you are applying in pairs, send in separate personal applications in which you state your preferred colleague. Selections will be made throughout the application period.
Supervisor Olof Steinert, YSFS
About the job
Title: Thesis Project in Machine Learning – Multilingual Large Scale Text Classification (MLSTC) for Troubleshooting Management
Business area: Research and Development
Last application date: Interviews in August 2021
Scania is a world-leading provider of transport solutions, including trucks and buses for heavy transport applications combined with an extensive product-related service offering. Scania offers vehicle financing, insurance and rental services to enable our customers to focus on their core business. Scania is also a leading provider of industrial and marine engines. In 2019, we delivered 91,700 trucks, 7,800 buses as well as 10,200 industrial and marine engines to our customers. Net sales totalled to over SEK 152 billion, of which about 20 percent were services-related. Founded in 1891, Scania now operates in more than 100 countries and employs some 51,000 people. Research and development are concentrated in Sweden, with branches in Brazil and India. Production takes place in Europe, Latin America and Asia, with regional production centers in Africa, Asia and Eurasia. Scania is part of TRATON SE.