Abstract: A server computing device, including memory storing a knowledge graph. The server computing device may further include a processor configured to receive a natural language input and generate a tokenized utterance based on the natural language input. Based on the tokenized utterance, the processor may select a predefined intention indicating a target ontology entity type of the natural language input. The processor may identify at least one input ontology entity token included in the tokenized utterance and may identify at least one relation between the predefined intention and the input ontology entity token. Based on the predefined intention, the at least one input ontology entity token, and the relation, the processor may generate a structured query. Based on the structured query and the knowledge graph, the processor may output an output ontology entity token having the target ontology entity type.

Abstract: A server computing device, including memory storing a knowledge graph including a plurality of ontology entities connected by a plurality of edges. The server computing device may further include a processor configured to generate a glossary file based on the knowledge graph. The glossary file may include a plurality of ontology entities included in the knowledge graph. The processor may receive a plurality of utterance templates. Each utterance template may include an utterance and a predefined intention. For each utterance template, the processor may generate one or more utterance template copies in which one or more ontology entities included in the utterance are replaced with one or more utterance template fields. The processor may generate a plurality of training utterances at least in part by filling the one or more utterance template fields of the one or more utterance template copies with respective ontology entities included in the glossary file.

Abstract: A server computing device, including memory storing a knowledge graph including a plurality of ontology entities. The server computing device may further include a processor configured to receive a tokenized utterance including a plurality of words and one or more metadata tokens. The processor may extract a respective word embedding vector from each word included in the tokenized utterance. Based on a glossary file, the processor may determine a respective ontology entity type of each word included in the tokenized utterance. The processor may extract a character embedding vector from each character included in the tokenized utterance. Based on the plurality of word embedding vectors, the plurality of respective ontology entity types of the words, and the plurality of character embedding vectors, the processor may determine a predefined intention of the tokenized utterance using at least one recurrent neural network. The predefined intention may indicate a target ontology entity type.

Abstract: A server computing device, including memory storing a knowledge graph. The server computing device may further include a processor configured to receive a natural language input and generate a tokenized utterance based on the natural language input. Based on the tokenized utterance, the processor may select a predefined intention indicating a target ontology entity type of the natural language input. The processor may identify at least one input ontology entity token included in the tokenized utterance and may identify at least one relation between the predefined intention and the input ontology entity token. Based on the predefined intention, the at least one input ontology entity token, and the relation, the processor may generate a structured query. Based on the structured query and the knowledge graph, the processor may output an output ontology entity token having the target ontology entity type.

Abstract: A server computing device, including memory storing a knowledge graph including a plurality of ontology entities. The server computing device may further include a processor configured to receive a tokenized utterance including a plurality of words and one or more metadata tokens. The processor may extract a respective word embedding vector from each word included in the tokenized utterance. Based on a glossary file, the processor may determine a respective ontology entity type of each word included in the tokenized utterance. The processor may extract a character embedding vector from each character included in the tokenized utterance. Based on the plurality of word embedding vectors, the plurality of respective ontology entity types of the words, and the plurality of character embedding vectors, the processor may determine a predefined intention of the tokenized utterance using at least one recurrent neural network. The predefined intention may indicate a target ontology entity type.

Abstract: A server computing device, including memory storing a knowledge graph including a plurality of ontology entities connected by a plurality of edges. The server computing device may further include a processor configured to generate a glossary file based on the knowledge graph. The glossary file may include a plurality of ontology entities included in the knowledge graph. The processor may receive a plurality of utterance templates. Each utterance template may include an utterance and a predefined intention. For each utterance template, the processor may generate one or more utterance template copies in which one or more ontology entities included in the utterance are replaced with one or more utterance template fields. The processor may generate a plurality of training utterances at least in part by filling the one or more utterance template fields of the one or more utterance template copies with respective ontology entities included in the glossary file.