Semantic entity relation detection classifier training

An apparatus comprising one or more computing devices, said computing devices being in communication with each other via a computer network whenever there is a plurality of computing devices, the apparatus comprising:

• a processor and memory storing instructions executable by the processor to implement
• a computer program having program modules to direct the one or more computing devices to,
• receive a query click log and a knowledge graph,
• find queries included in the query click log that are associated with entities found in the knowledge graph, said entities being associated with a knowledge graph domain of interest,
• use the knowledge graph to identify implicit relations from the found queries,
• based on identifying the implicit relations from the found queries, generate an implicit relations data set comprising queries associated with the identified implicit relations, and
• train a semantic entity relation detection classifier using at least the implicit data set to find a relation in a query, said found relation being utilized to automatically generate an appropriate response to the query whenever it is input to said trained classifier.

An apparatus comprising one or more computing devices, said computing devices being in communication with each other via a computer network whenever there is a plurality of computing devices, the apparatus comprising:

• a processor and memory storing instructions executable by the processor to implement
• a computer program having program modules to direct the one or more computing devices to,
• receive a query click log and a knowledge graph,
• find queries included in the query click log that are associated with entities found in the knowledge graph, said entities being associated with a knowledge graph domain of interest,
• use the knowledge graph to identify explicit relations from the found queries,
• based on identifying the explicit relations from the found queries, generate an explicit relations data set comprising queries associated with the identified explicit relations,
• use the knowledge graph to identify implicit relations from the found queries,
• based on identifying the implicit relations from the found queries, generate an implicit relations data set comprising queries associated with the identified implicit relations,
• train a first classifier using the implicit relations data set to produce an implicit relations classifier that can find implicit relations in a query,
• apply the implicit relations classifier to each of the queries in the explicit relations data set to find queries predicted to have an implicit relation or implicit relations,
• augment the explicit relations data set, said augmenting comprising, for each query in the explicit relations data set predicted to have the implicit relation or implicit relations, adding the implicit relation or implicit relations predicted for that query to the explicit relations data set entry associated with the query to produce an augmented explicit relations data set, and
• train a second classifier using the augmented explicit relations data set to produce a combined relations classifier that can find explicit, or implicit relations, or both, in the query, the found explicit or implicit relations, or the found explicit and implicit relations, being utilized to automatically generate an appropriate response to the query whenever it is input to the combined relations classifier.