Take sound adjustment as an example. Assuming that the intention of sound adjustment is determined by the three word slots of "operation object executive email list", "adjustment" and "status", the "sound", "raise" and "a little bit" in the sentence "raise the sound a little" correspond to the "operation object" respectively. ", "adjustment" and "status" executive email list three word slots. If the user says "please help me raise my voice a little bit", and the intent cannot be matched due to the addition of the words "please help me", then the intent matching rule of the system is a complete match.
If the match is successful, it means that the intent matches. Rules support fuzzy matching. The voice intelligence platform that only supports the complete executive email list matching of word slots has almost no generalization ability. At this time, designers executive email list need to consider the generalization of intentions by building dictionaries, word slots and slots, which will test the designer's language understanding level and logic. The ability and executive email list the overall design ability of the overall dictionary, word slot, slot, we can think that this task is extremely difficult.
If the voice intelligence platform supports fuzzy matching of word slots, it means that the system has adopted the method of identifying keywords. Taking the executive email list example of "please improve my voice a little bit" just now, the system can identify that "improving my voice a little bit" belongs to the "operation object", The three word slots "adjustment" and "status" will then match the corresponding intent, while other words "please help me" or "please help executive email list me" will be ignored. Fuzzy matching ability can significantly improve the generalization ability of intent, which can greatly reduce the workload of designers, because we try to choose a voice intelligent platform with fuzzy matching ability as much as possible . 3.