"Content" is the fundamental element in the digital advertising eco-system which connects users, publishers, and advertisers. With advanced algorithms, we can analyze and understand a variety of content such as Web documents, search queries, images, and videos. Precise analysis of Web and user generated content helps us understand user intent, which enables the delivery of relevant advertisements and information to the users. Our technologies in this research area have been adopted in Microsoft's Content Ads and Paid Search services. In addition, we continue to invest in developing new, state-of-the art algorithms and technologies in this area to improve user experiences and help our partners succeed in reaching the right audience.
A brief introduction of the tools in this category is provided below. You can explore more by clicking on each individual tool.
Location Platform
This tool shows a few key Location Intelligence technologies. Based on the user input, it determines a few location-based aspects of user's intent, including "what", "where" (target location), local intent and travel intent. It then displays the target location on the map and also provides the location-based or topic-based query suggestions.
Visual Product Browsing
This tool uses computer vision algorithms to browse and categorize images as a human might, without the need for manual data tagging. Online shoppers could use this technology to visually browse categorized product images (such as lamps) and narrow their search to find 'more like this' options.
People Name Detection
Many online queries are either personal names or they contain personal names. For example,
"John Smith" is a personal name, and
"I am pleased to tell you that Andy Beal will be working as an internet marketing consultant" contains a personal name. This tool detects names of people in a query, which can improve ad relevancy for online marketing. This also helps in understanding a user's Web search intent and could be used to provide more relevant search results to meet the user's needs, such as providing the biography and other attributes of the targeted person to the users.
Product Classification
This tool classifies products, and queries about products, into the MSN shopping taxonomy. It can help merchants understand what kind of products customers are searching for on their web site. This allows merchants to provide better services to meet users' needs, thus driving customer satisfaction and generating profits simultaneously.
Cross-Language Keyword Suggestion
This tool suggests relevant queries in a different language based on the query logs. Instead of performing an exact translation, it suggests relevant and popular queries of the target language, and enables various existing keyword technologies, such as bid term suggestion, SEO, query suggestion, to cross the language boundaries. For example, it can translate UNEF (United Nations Emergency Force) to FUNU (Force Urgence des Nations Unies). This tool may help advertiser to quickly identify relevant foreign language bid terms, facilitating targeted advertising in the multi-lingual scenario.
Content Categorization
This tool displays the categories most related to a Web page based on its content. It also shows the degree of confidence that a category relates to the page's content.
Entity Analysis
This tool analyzes the query entered by a user and breaks it into individual entity names. It can detect a multiple-word entity. For example, for the query "restaurants near space needle," the tool will recognize that "space needle" is an unbreakable entity name. If a query contains more than one entity, this tool successively analyzes each entity. This tool is useful when performing a match between keywords and search queries, because it is important that entities should not be broken during the matching process.
Sensitive Webpage Detection
This tool helps determine whether a Web page contains content that might be considered sensitive by the general public, such as content about porn, terrorism, or crime.
Context-Based Acronym Resolution
When a user enters an acronym or abbreviation, this tool predicts the most likely result based on the acronym's context. For example, if the query is ms cure, this tool offers multiple sclerosis as the most likely result, as well as Microsoft, millisecond, and other alternatives.
Ad Text Writer
This tool generates possible ads for web site entered by a user, based on the keywords that are most representative of content on that web site. This tool is useful to automatically generate multiple ads for large web sites that have many keywords.