Abstract: Function expansion is an important way to achieve innovative product design, and data-drivenis also the key supporting technology for product function expansion. In this context, a product function expansion method driven by text mining and patent data collaboration is proposed and expected to provide a quantitative research method for changing and upgrading existing products. Firstly, the function-structure mapping information of the target product are analyzed, then a partial hypergraph model based on existing network models is proposed to calculate the weight of product's components. Secondly, four extracting product functional and structural knowledge, text data of us patent is retrieved and obtained, then part of speech tagging (POS tagging) and dependency parsing (DP) are used to construct mining rules to extract subject-action-object (SAO) in sentences. Thirdly, patent text is used as the corpus and BERT(Bidirectional encoder representations from transformers) model is utilized to train word vectors as a tool for finding similar structures and functions, then four function expansion strategies combined with collaborative recommendationalgorithms are proposed as follows: other function expansion, similarity function expansion, similar component’s functions expansion, and similar components’ functions expansion based on the target product components; finally, according to the requirements of enterprise, the case of showerhead is applied to verifies thefeasibility and effectiveness of the method.

Key words: function expansion, collaborative recommendation, patent text, partial hypergraph, BERT