com.cloudera.oryx.app.als

Interface RescorerProvider

All Known Implementing Classes:

AbstractRescorerProvider, MultiRescorerProvider

public interface RescorerProvider

Implementations of this interface provide, optionally, objects that can be used to modify and influence the results of endpoints like Recommend, Similarity, etc. It is a means to inject business logic into the results of the ALS application endpoints.

Implementations of this class are factories. An implementation creates and configures a Rescorer rescoring object and returns it for use in the context of one of the endpoints' operation. For example, the result of getRecommendRescorer(List, List) is used with Recommend. The Rescorer filters the candidates recommendations by item ID (Rescorer.isFiltered(String)) or modifies the scores of item candidates that are not filtered (Rescorer.rescore(String, double)) based on the item ID and original score.

The factory methods take several optional String arguments. These are passed from the REST API, as several Strings, from URL parameter rescorerParams. The implementation may need this information to initialize its rescoring logic for the request. For example, the argument may be the user's current location, used to filter results by location.

For example, a request containing ...?rescorerParams=xxx,yyy,zzz will result in an args parameter with one elements, xxx,yyy,zzz. A request containing ...?rescorerParams=xxx&...rescorerParams=yyy&...rescorerParams=zzz... will result in an args parameter with 3 elements, xxx, yyy, zzz.

Since:

2.0.0

Method Summary

Modifier and Type	Method and Description
Rescorer	getMostActiveUsersRescorer(List<String> args)
Rescorer	getMostPopularItemsRescorer(List<String> args)
Rescorer	getMostSimilarItemsRescorer(List<String> args)
Rescorer	getRecommendRescorer(List<String> userIDs, List<String> args)
Rescorer	getRecommendToAnonymousRescorer(List<String> itemIDs, List<String> args)

Method Detail

getRecommendRescorer

Rescorer getRecommendRescorer(List<String> userIDs,
                              List<String> args)

Parameters:

userIDs - user(s) for which recommendations are being made, which may be needed in the rescoring logic.

args - arguments, if any, that should be used when making the Rescorer. This is additional information from the request that may be necessary to its logic, like current location. What it means is up to the implementation.

Returns:

Rescorer to use or null if none should be used. The resulting Rescorer will be passed each candidate item ID to Rescorer.isFiltered(String), and each non-filtered candidate with its original score to Rescorer.rescore(String, double)

Since:

2.0.0

getRecommendToAnonymousRescorer

Rescorer getRecommendToAnonymousRescorer(List<String> itemIDs,
                                         List<String> args)

Parameters:

itemIDs - items that the anonymous user is associated to

args - arguments, if any, that should be used when making the Rescorer. This is additional information from the request that may be necessary to its logic, like current location. What it means is up to the implementation.

Returns:

Rescorer to use or null if none should be used. The resulting Rescorer will be passed each candidate item ID to Rescorer.isFiltered(String), and each non-filtered candidate with its original score to Rescorer.rescore(String, double)

Since:

2.0.0

getMostPopularItemsRescorer

Rescorer getMostPopularItemsRescorer(List<String> args)

Parameters:

args - arguments, if any, that should be used when making the Rescorer. This is additional information from the request that may be necessary to its logic, like current location. What it means is up to the implementation.

Returns:

Rescorer to use or null if none should be used. Note that only its Rescorer.isFiltered(String) method has effect.

Since:

2.0.0

getMostActiveUsersRescorer

Rescorer getMostActiveUsersRescorer(List<String> args)

Parameters:

args - arguments, if any, that should be used when making the Rescorer. This is additional information from the request that may be necessary to its logic, like current location. What it means is up to the implementation.

Returns:

Rescorer to use or null if none should be used. Note that only its Rescorer.isFiltered(String) method has effect.

Since:

2.0.0

getMostSimilarItemsRescorer

Rescorer getMostSimilarItemsRescorer(List<String> args)

Parameters:

args - arguments, if any, that should be used when making the Rescorer. This is additional information from the request that may be necessary to its logic, like current location. What it means is up to the implementation.

Returns:

Rescorer to use or null if none should be used. The Rescorer will be passed, to its Rescorer.isFiltered(String) method, the candidate item ID passed in the user query as its second element. Each non-filtered pair is passed with its original score to Rescorer.rescore(String, double)

Since:

2.0.0