Computational Advertising Based on Text Analysis
In recent years, text analysis has become a research hotspot in the field of natural language processing. Some applied researchers in this discipline have successively applied the research results of web page analysis, text tendency analysis, text similarity analysis and machine translation to online advertising practice, among which Baidu bidding ranking, Google AdSense and DoubleClick contextual advertising are outstanding representatives. Although they are all based on text analysis, and they all have a common premise that users like advertisements related to their own information needs, their system principles are completely different. Baidu bidding ranking can insert advertisements related to users' search in front of search results by calculating the similarity between users' search keywords and advertisers' bid keywords. Obviously, Baidu's bidding ranking is suspected of misleading users through human intervention in search results. In order to ensure the fairness of search results, Google gave up bidding advertisements earlier, and instead put advertisements similar to search terms on the search target page in the form of webmaster joining, that is, Google AdSense. On the other hand, Contextual Ad searches for advertisements with similar themes from the advertisement library based on the theme analysis of the webpages browsed by users, and inserts them into the designated positions of webpages to realize the matching of advertisements and contexts.
Computational Advertising Based on User Analysis
If the matching between advertisement and context based on content analysis indirectly realizes the matching between advertisement and user, then computing advertisement based on user analysis directly seeks the consistency between advertisement and user. At present, user analysis mainly starts from IP, registration data, server logs, Cookie, historical data, browser behavior and other aspects, and its representative advertising forms include personalized recommendation advertisements of e-commerce and MediaV. Personalized recommendation advertisement can be regarded as an intelligent improvement of POP advertisement, which recommends products that users are interested in according to their interest characteristics and purchase behavior. This positioning process combines IP state, information retrieval, collaborative filtering, data mining and other algorithms. MediaV tracks users' browsing history and analyzes users' interest orientation according to Cookie. When users log on to a web page, they can identify users through the MediaV platform and put in advertisements that meet their interests.
Computational Advertising Based on User Participation
Both text and user behavior can be analyzed by association algorithm. However, under the existing image recognition technology, multimedia data such as images and videos cannot be subject analyzed, and manual participation is needed. The main purpose of computing advertising system based on user participation is to build an alliance platform for users, advertisers and webmasters, such as advertising alliance in the picture of Pixazza and advertising alliance in the video of Qiyi. The alliance advertising system attracts netizens to participate in advertising creation activities as volunteer experts in the form of profit sharing. When experts browse pictures or videos and find commodity information, they insert advertising interest points in corresponding positions, link to corresponding commodities and buy advertisements. Different from traditional video advertising, this advertising system only displays advertisements when users point to the interest points of advertisements, which reduces the disturbance to users in browsing pictures and videos to some extent.