Annual report-oriented company name and relationship extraction method

The invention discloses an annual report-oriented company name and relationship extraction method and discloses an annual report-oriented company name and relationship extraction method. The method comprises the following specific steps: 1, training a word vector model; 2, defining the classification of the company relationship, setting labels of SVM classification into three types, namely a non-investment relationship, an investment relationship and a non-relationship respectively; 3, manually screening sentences reflecting the company relationship, and 4, segmenting the sentences in the company relationship text corpus through jaeba, and converting the segmented sentences into vectors through a WORD2VEC word vector model to be expressed, and relates to the technical field of relationshipextraction. According to the annual report-oriented company name and relationship extraction method, the problem that a widely used template cannot be expanded from a set of relationship templates due to the fact that expert