Method for performing approximate search based on word vectors to quickly extract advertisement text themes

Abstract

The invention discloses a method for carrying out approximate search based on word vectors to rapidly extract advertisement text themes. The method comprises the following steps of: step 1, searchingwords which are the same as words in a stop word bank from an advertisement title by using a vocabulary segmentation tool and the existing stop word bank, removing the words, namely, stop words in theadvertisement title, extracting Chinese words in a corpus, taking the Chinese words as a dictionary, and performing word segmentation on an advertisement text theme by using the dictionary. The method is convenient to operate,. The search complexity of a single query word in the GPU-DMM generation model can be reduced from 0 (N) to 0 (log N); the whole advertisement text topic extraction processis accelerated, the extraction speed is greatly increased, offline processing and unsupervised training can be completed in several hours in the whole process, the requirements for large-scale data volume and near real-time performance of the internet advertisement industry can be met, and user interest tags can be updated day by day or hour by hour.

Description

technical field [0001] The invention relates to a method for extracting an advertisement text theme, in particular to a method for quickly extracting an advertisement text theme by performing approximate search based on a word vector. Background technique [0002] In the Internet advertisement recommendation business, firstly, the subject of the advertisement text is extracted according to the advertisement text clicked or browsed by the user, and then the user’s interest tag is determined. Common ones are LDA and GPU-DMM. [0003] LDA is a document topic generation model, which contains a three-layer structure of words, topics and documents. The so-called generative model means that we believe that each word in an article is obtained through the process of "selecting a certain topic with a certain probability, and selecting a certain word from this topic with a certain probability". Documents to topics obey a multinomial distribution, and topics to words obey a multinomia...

AI Technical Summary

Problems solved by technology

[0013] Aiming at the deficiencies in the background technology, the present invention designs a method for quickly extracting the subject of advertisement text by performing approximate search based on word vectors, the purpose of which is to solve the problem of slow extraction of subject of advertisement text in the prior art

Images