Microsoft natural language search


















Feedback will be sent to Microsoft: By pressing the submit button, your feedback will be used to improve Microsoft products and services. Privacy policy. All Microsoft Academic Services have been officially retired as of December 31st, We are currently retaining original documentation as-is for educational use, however all information about signing up for services is no longer valid, and support and service API links will not function.

This article describes how MAKES generates semantic interpretations of a natural language query that enable entities to be retrieved. It is intended for developers using MAKES who require a deeper understanding of the architecture and techniques used. Processing a natural language query is a multi-stage process that starts with lexical analysis of the query, allowing the query to be rewritten for optimal search performance. Valid hypotheses are resolved into full semantic query expressions which are in turn used to generate the complete interpretation response.

Lexical analysis of the user query is done to help ensure that valid interpretation hypotheses can be generated. MAKES supports a small handful of query operators that tell the grammar parser to perform a specific way when generating hypotheses:. In this example, the Excel workbook with data about summer Olympics has been saved to Office Similarly this sentence is being interpreted to count the number of distinct athletes that won medals at the last London Olympics.

One can easily direct the system to display a different visualization. It opted to display maps because the country column is marked as a geography data column. When using natural language there are often cases where the question posed contains significant ambiguity.

For example:. The challenges our team faces stem from the highly ambiguous nature of natural language. Yet this sentence presents difficulties to a software program because it is ambiguous and relies on real-world knowledge. How much and what sort of context needs to be brought to bear on these questions in order to adequately disambiguate the sentence?

Our work has implications for applications such as text critiquing, information retrieval, question answering, summarization, gaming, and translation. For example, the grammar checkers in Office for English, French, German, and Spanish are outgrowths of our research.



0コメント

  • 1000 / 1000