Skip to content

Latest commit

 

History

History
87 lines (75 loc) · 3.14 KB

evaluateMethod.md

File metadata and controls

87 lines (75 loc) · 3.14 KB
Raw

evaluate Method

The evaluate method evaluates and returns the output of a structured query expression based on the index data.

Typically, an expression will be obtained from a response to the interpret method. But you can also compose query expressions yourself (see Structured Query Expression).

Request

http://<host>/evaluate?expr=<expr>&attributes=<attrs>[&<options>]

Name Value Description
expr Text string Structured query expression that selects a subset of index entities.
attributes Text string Comma-delimited list of attributes to include in response.
count Number (default=10) Maximum number of results to return.
offset Number (default=0) Index of the first result to return.
orderby Text string Name of attribute used to sort the results, followed by optional sort order (default=asc): "attrname[:(asc|desc)]". If not specified, the results are returned by decreasing natural log probability.
timeout Number (default=1000) Timeout in milliseconds. Only results computed before the timeout has elapsed are returned.

Using the count and offset parameters, a large number of results may be obtained incrementally over multiple requests.

Response (JSON)

JSONPath Description
$.expr expr parameter from the request.
$.entities Array of 0 or more object entities matching the structured query expression.
$.aborted True if the request timed out.

Each entity contains a logprob value and the values of the requested attributes.

Example

In the academic publications example, the following request passes a structured query expression (potentially from the output of an interpret request) and retrieves a few attributes for the top 2 matching entities:

http://<host>/evaluate?expr=Composite(Author.Name=='jaime teevan')&attributes=Title,Y,Author.Name,Author.Id&count=2

The response contains the top 2 ("count=2") most likely matching entities. For each entity, the title, year, author name, and author ID attributes are returned. Note how the structure of composite attribute values matches the way they are specified in the data file.

{
  "expr": "Composite(Author.Name=='jaime teevan')",
  "entities": 
  [
    {
      "logprob": -6.645,
      "Ti": "personalizing search via automated analysis of interests and activities",
      "Y": 2005,
      "Author": [
        {
          "Name": "jaime teevan",
          "Id": 1968481722
        },
        {
          "Name": "susan t dumais",
          "Id": 676500258
        },
        {
          "Name": "eric horvitz",
          "Id": 1470530979
        }
      ]
    },
    {
      "logprob": -6.764,
      "Ti": "the perfect search engine is not enough a study of orienteering behavior in directed search",
      "Y": 2004,
      "Author": [
        {
          "Name": "jaime teevan",
          "Id": 1982462162
        },
        {
          "Name": "christine alvarado",
          "Id": 2163512453
        },
        {
          "Name": "mark s ackerman",
          "Id": 2055132526
        },
        {
          "Name": "david r karger",
          "Id": 2012534293
        }
      ]
    }
  ]
}