Natural Language Processing System

Abstract

A natural language processing system includes a processing unit and a storage unit electrically connected to the processing unit. The processing unit is configured to: upon receiving a query message expressing a question, determine whether one of a plurality of priority references exists that semantically matches the query message; if yes, generate a first natural language response message based on the query message and the priority reference, using the natural language processing model, and output the response message; if no, generate a second natural language response message using the natural language processing model, following response rules indicated by a set of normative data, and output the second natural language response message.

Claims

1. A natural language processing system suitable for use with a communication device; the natural language processing system comprising: a processing unit suitable for electrical connection with the communication device; and a storage unit electrically connected to the processing unit, and storing a reference data set containing multiple priority references, and a standard data set, wherein, The specification data set indicates the response specifications that a natural language processing model must follow when generating natural language response messages. The processing unit is configured to: upon receiving a query message corresponding to the communication device and expressing a consultation question, use the natural language processing model to determine whether there is a matching priority reference in the priority references of the reference data set that semantically matches the query message; if a matching priority reference is found, generate a first natural language response message to the consultation question using the natural language processing model based on the query message and the matching priority reference, and provide the first natural language response message to the communication device; and if no matching priority reference is found, generate a second natural language response message to the consultation question using the natural language processing model based on the query message and the specification data set, following the response specifications indicated by the specification data set, and provide the second natural language response message to the communication device.

2. The natural language processing system as described in claim 1, wherein: The storage unit also stores a network search module; the response specifications indicated by the specification data set include a network trigger condition corresponding to a specific query. The processing unit is further configured to: when it is determined that there are no matching priority references among the priority references, use the natural language processing model to determine whether the connection triggering condition is met based on the query message, wherein the connection triggering condition includes that the consultation question expressed in the query message matches the semantics of the specific inquiry; when it is determined that the connection triggering condition is met, control the natural language processing model to call the internet search module to perform an internet search process on the consultation question, and use the natural language processing model to generate a third natural language response message to the consultation question based on an internet search result obtained through the internet search process, and provide the third natural language response message to the communication device; and the processing unit only generates the second natural language response message using the natural language processing model when there are no matching priority references among the priority references and when it is determined that the connection triggering condition is not met, and provides the second natural language response message to the communication device.

3. The natural language processing system as described in claim 1, wherein: Each set of priority references includes a reference question and a reference answer to the reference question; the processing unit determines whether there is a matching priority reference among the priority references by determining whether the reference question contained in one of the priority references matches the semantics of the inquiry question expressed in the question message, so as to determine whether there is a matching priority reference among the priority references; and the processing unit generates the first natural language response message by using the natural language processing model based on the question message and the reference answer contained in the matching priority reference.

4. The natural language processing system as described in claim 1, wherein: The specification data set includes a knowledge base portion, which includes multiple supplementary references; the response specification indicated by the specification data set includes a data boundary rule indicating the knowledge base portion; the processing unit generates the second natural language response message using the natural language processing model by: selecting one or more matching supplementary references from the supplementary references in the knowledge base portion that have the highest semantic matching degree with the question message, and restricting the reference source when generating the second natural language response message using the natural language processing model to the one or more matching supplementary references according to the data boundary rule.

5. The natural language processing system as described in claim 2, wherein: The response specifications indicated by the set of specifications include a topic compliance condition, wherein the topic compliance condition indicates a pre-determined specific consultation topic; the processing unit is further configured to: when it is determined that there is no matching priority reference among the priority references, use the natural language processing model to determine whether the query message meets the topic compliance condition, wherein the topic compliance condition includes that the consultation question expressed by the query message belongs to the scope of the specific consultation topic; when it is determined that the query message does not meet the topic compliance condition, use the natural language processing model to generate a rejection response message that does not respond to the consultation question, and provide the rejection response message to the communication device; and the processing unit only determines whether the connection trigger condition is met based on the query message when there is no matching priority reference among the priority references and the query message meets the topic compliance condition.

6. A natural language processing system, suitable for use with a communication device and a voice service system; the natural language processing system comprising: a processing unit, suitable for electrical connection with the communication device and the voice service system; and a storage unit, electrically connected to the processing unit, and storing multiple sets of reference data and multiple sets of specification data, wherein, Each set of reference data indicates the response specifications that a natural language processing model must follow when generating natural language response messages; wherein the processing unit is configured to: upon receiving a telecommunications call request transmitted by the voice service system and containing a service telephone number, select one reference data set corresponding to the service telephone number from the set of reference data, and select one set of reference data corresponding to the service telephone number from the set of reference data; approve the telecommunications call request so that the voice service system establishes a voice communication channel between the communication device and the natural language processing system; and during communication with the communication device through the voice communication channel, control the natural language processing model to generate natural language based on the selected reference data set and the set of reference data.

Images