Knowledge crowdsourcing cold start task modeling and recommendation method

Abstract

The invention discloses a knowledge crowdsourcing cold start task modeling and recommendation method, and relates to the technical field of knowledge crowdsourcing cold start task recommendation, and the method comprises the following steps: S1, obtaining external data, and constructing a training set and a test set; s2, constructing an interaction prediction model, wherein the interaction prediction model comprises a self-attention mechanism network layer; and S3, inputting the explicit attribute of the knowledge crowdsourcing cold start task to be predicted into the trained interaction prediction model. According to the method, explicit attributes of the cold start task, the similar historical task, the service provider and the historical participation task are introduced; a self-attention mechanism is adopted to respectively capture a coupling relationship between a knowledge crowdsourcing cold start task and similar historical task features thereof and between a service provider and participated historical task features thereof, attention weights are distributed, fusion vector representation of the knowledge crowdsourcing cold start task and the service provider can be effectively and accurately learned, and a new thought is provided for solving task cold start, relieving information overload and realizing task-service provider effective matching for knowledge crowdsourcing.

Description

technical field [0001] The invention relates to the technical field of crowdsourcing recommendation, in particular to a knowledge crowdsourcing cold start task modeling and recommendation method. Background technique [0002] Knowledge crowdsourcing is a new type of knowledge economy model that uses the platform as the medium and uses the professional knowledge and skills of service providers to solve the business development and innovation problems of employers. With the rise of the Internet and gig economy, more and more organizations and individuals choose to join knowledge crowdsourcing platforms. However, due to task diversity, service provider personalization, and platform network externalities, the problem of information overload on knowledge crowdsourcing platforms has become increasingly prominent, and it has become increasingly difficult to achieve effective matching between employers and service providers. On the one hand, it is difficult for employers to hire sui...

AI Technical Summary

Problems solved by technology

However, there may be problems such as the lack of knowledge crowdsourcing task attributes in real scenarios; and because the attributes of knowledge crowdsourcing tasks are mostly described by natural language, there may be inaccurate descriptions, and the similarity calculated based on this may also be inaccurate

In addition, service providers' historical participation in knowledge crowdsourcing tasks is not equally important in characterizing their personalized behavior, which is not considered by such methods

Images