Microsoft Technology Licensing, LLC

Shibnath Mukherjee

Manish Chandra Joshi

Rupam Basu

Shiva Manohar Singh

Ranjita Jatti Naik

Embodiments of the disclosure extract annotations from web pages. The annotations are combined with search results and/or advertisements to allow the user to better understand the content of the search result or advertisement landing web page. A visual snapshot of the web page is taken. Visual processing extracts information from the visual representation. The HTML, for the web page is also analyzed and various pieces of information extracted. The information from the visual processing is combined with the information extracted from the HTML. The combined information is evaluated and information for the annotations are selected. The annotations are then combined with the search results and/or advertisements.

Visual parsing for annotation extraction

Embodiments of the disclosure extract annotations from web pages. The annotations are combined with search results and/or advertisements to allow the user to better understand the content of the search result or advertisement landing web page. A visual snapshot of the web page is taken. Visual processing extracts information from the visual representation. The HTML for the web page is also analyzed and various pieces of information extracted. The information from the visual processing is combined with the information extracted from the HTML. The combined information is evaluated and information for the annotations are selected. The annotations are then combined with the search results and/or advertisements.

Methods, systems, and computer-readable storage media for predicting advertiser keyword performance indicator values based on proximity of advertisers to one another are provided. A keyword is received from a first advertiser that has not previously bid on the keyword. Another advertiser that is similar to the first advertiser and has previously bid on the keyword is identified, and a predicted performance indicator value for the keyword is calculated for the first advertiser based upon an established performance indicator value of the keyword for the other advertiser. In this way, performance predictions for an advertiser desiring to bid on a keyword that is new to that advertiser may be calculated more accurately than traditional methods that do not take into account advertiser similarity.

Growing community of inventors

Bengaluru, India

Shibnath Mukherjee