Method and apparatus for identifying key phrase in audio, device and medium

A method for identifying a key phrase in audio, comprising:

• obtaining audio data to be identified;
• and identifying the key phrase in the audio data by using a trained key phrase identification model as a neural network, wherein the key phrase identification model is trained based on first training data for identifying feature information of words in a first training text and second training data for identifying the key phrase in a second training text, wherein obtaining the first training data comprises:
• obtaining the first training text;
• splitting the first training text into at least one sentence;
• and determining the feature information of the words in the at least one sentence using a natural language processing technology, and wherein obtaining the second training data comprises:
• obtaining the second training text;
• splitting the second training text into at least one sentence;
• and identifying the key phrase in the at least one sentence using a plurality of labels, wherein identifying the key phrase in the audio data by using the trained key phrase identification model comprises:
• determining a corresponding label fora character in each sentence in a text converted from the audio data by using the key phrase identification model, and identifying the key phrases in the sentence based on the corresponding label, wherein identifying the key phrases in the sentence based on the corresponding label comprises:
• in response to identifying a starting character identified by a first label, a subsequent character identified by a second label, and a character identified by a third label in the at least one sentence that does not belong to the key phrase, identifying a set consisting of the starting character identified by the first label and the subsequent character identified by the second label as the key phrase.

An apparatus for training a key phrase identification model, comprising:

• one or more processors, and a storage device, configured to store one or more programs, wherein, when the one or more programs are executed by the one or more processors, the one or more processors are configured to implement a method for training a key phrase identification model, comprising:
• obtaining first training data for identifying feature information of words in a first training text;
• obtaining second training data for identifying a key phrase in a second training text;
• and training the key phrase identification model to be used as a neural network, based on the first training data and the second training data, so as to identify the key phrase in audio data by using the trained key phrase identification model, wherein obtaining the first training data comprises:
• obtaining the first training text;
• splitting the first training text into at least one sentence;
• and determining the feature information of the words in the at least one sentence using a natural language processing technology, and wherein obtaining the second training data comprises:
• obtaining the second training text;
• splitting the second training text into at least one sentence;
• and identifying the key phrase in the at least one sentence using a plurality of labels, wherein identifying the key phrase in the audio data by using the trained key phrase identification model comprises:
• determining a corresponding label for a character in each sentence in a text converted from the audio data by using the key phrase identification model, and identifying the key phrases in the sentence based on the corresponding label, wherein identifying the key phrases in the sentence based on the corresponding label comprises:
• in response to identifying a starting character identified by a first label, a subsequent character identified by a second label, and a character identified by a third label in the at least one sentence that does not belong to the key phrase, identifying a set consisting of the starting character identified by the first label and the subsequent character identified by the second label as the key phrase.