TWI703456B - Intelligent recommendation system and intelligent recommendation method - Google Patents

Abstract

An intelligent recommendation system includes a voice segmentation circuit, a speech recognition circuit, a plurality of category databases, and a processing circuit. The voice segmentation circuit is configured to receive speech data and segment it according to the content of the voice data. The speech recognition circuit is configured to recognize the semantics of the speech data according to the segmented speech data. Each of the plurality of category databases is configured to store a plurality of answers. The processing circuit is configured to search for the answers related to the semantic meaning of the speech data in the category database according to the semantic meaning of the speech data. The processing circuit sorts the answers according to the relevance between the semantic meaning of the speech data and the answer, and recommends a recommended answer in the sorted answer.

Description

Intelligent recommendation system and method

This disclosure relates to an intelligent recommendation system and an intelligent recommendation method, in particular an intelligent recommendation system and an intelligent recommendation method capable of intelligent question and answer.

With the development of artificial intelligence, voice recognition technology has become more and more mature and has been widely used in many different fields, such as smart assistants for mobile devices, smart home appliances, and so on. Voice recognition technology has brought a lot of convenience and help to people’s lives, but traditional voice recognition technology is limited to a single field, or only provides links for users’ questions, users must click on the link to find out if there is an answer. Users cannot get answers directly by asking questions.

In an embodiment of the present disclosure, an intelligent recommendation system includes a voice segmentation circuit, a voice recognition circuit, a plurality of category databases, and a processing circuit. The voice segmentation circuit is used to receive the voice data and perform segmentation according to the content of the voice data. The voice recognition circuit is used to base the segmented voice data Identify the semantic meaning of the voice data. The plural category databases are each used to store plural answers. The processing circuit is used to search for answers related to the semantic meaning of the voice data in the category database according to the semantics of the voice data. The processing circuit sorts the answers according to the semantics of the voice data and the relevance of the answers, and recommends the sorted answers Recommended answer in answer.

In another embodiment of the present disclosure, an intelligent recommendation method includes the following operations: receiving voice data through a voice segmentation circuit, and performing segmentation based on the content of the voice data; using a voice recognition circuit, identifying based on the segmented voice data Find out the semantic meaning of the voice data; according to the semantic meaning of the voice data, search for multiple answers related to the semantic meaning of the voice data in the plural category databases; analyze the relevance between the semantic meaning of the voice data and the answers, and sort the answers ; Recommend the recommended answer of the ranked answer.

In summary, by recognizing the semantic meaning of the voice data and searching for multiple answers related to the semantic meaning of the voice data in a plurality of category databases, searching for the most corresponding answer, the user’s problem can be solved intelligently.

100‧‧‧Smart Recommendation System

110‧‧‧Speech segmentation circuit

120‧‧‧Voice recognition circuit

130‧‧‧Processing circuit

140-1, 140-2, 140-n‧‧‧Category database

200‧‧‧Method

S210, S220, S230, S240, S250‧‧‧Step

Fig. 1 is a functional block diagram of an intelligent recommendation system according to an embodiment of the present disclosure.

Figure 2 shows a flowchart of a smart recommendation method according to an embodiment of the present disclosure.

The terms "include", "have" and so on used in this article are all open terms, meaning "including but not limited to". In addition, the "and/or" used in this article includes any one of one or more of the related listed items and all combinations thereof.

In this text, when an element is referred to as "connection" or "coupling", it can refer to "electrical connection" or "electrical coupling". "Link" or "coupling" can also be used to indicate mutual operation or interaction between two or more components. In addition, although terms such as "first", "second", ... are used herein to describe different elements, the terms are only used to distinguish elements or operations described in the same technical terms. Unless the context clearly indicates, the terms do not specifically refer to or imply the order or sequence, nor are they used to limit this disclosure.

Please refer to FIG. 1. FIG. 1 is a functional block diagram of a smart recommendation system according to an embodiment of the present disclosure. The intelligent recommendation system 100 includes a voice segmentation circuit 110, a voice recognition circuit 120, a processing circuit 130, a category database 140-1, a category database 140-2, ..., a category database 140-n.

In one embodiment, the smart recommendation system 100 may be a wearable electronic device, a mobile electronic device, or other types of electronic devices, and the voice segmentation circuit 110 is used to receive voice messages sent by the user, such as asking questions. The voice recognition circuit 120 is used to recognize the semantic meaning of the voice message. The processing circuit 130 analyzes the semantic meaning of the voice message and searches and sorts the data in the category database 140-1, category database 140-2,..., and category database 140-n, and displays the answer corresponding to the voice message to the user By. In this example, the class The number of other databases is n as an example, and the number of n can be adjusted according to actual applications.

It should be noted that the implementation of the devices and components in the above-mentioned smart recommendation system 100 is not limited to those disclosed in the above-mentioned embodiments, and the connection relationship is not limited to the above-mentioned embodiments. Anything that is sufficient to enable the smart recommendation system 100 to implement the following Both the connection method and the realization method of the technical content can be applied to this case.

Figure 2 shows a flowchart of a smart recommendation method according to an embodiment of the present disclosure. The method 200 in FIG. 2 includes step S210, step S220, step S230, step S240, and step S250. In order to make the smart recommendation method shown in Figure 2 easy to understand, please refer to Figure 1 and Figure 2 at the same time. In step S210, the user's voice data, such as a user's question, is received through the voice segmentation circuit 110, and the voice segmentation circuit 110 will segment the voice data after receiving the voice data. In one embodiment, the user asks "Is it possible to buy stocks at the age of 18", and the voice segmentation circuit 110 divides "18 years old" into the first segment, "Can" be divided into the second segment, and "Buy stocks" is divided into the third segment. segment.

In step S220, the voice recognition circuit 120 recognizes the semantic meaning of the voice data from the segmented voice data. In one embodiment, the voice recognition circuit 120 converts the segmented voice data into a voice vector, for example, converts the three paragraphs of "18 years old", "can it" and "buy stock" into a combination of 0 and 1. Then, compare these three speech vectors with n category database 140-1, category database 140-2, ..., category database 140-n. The category database can be, for example, encyclopedia module, solitaire game module, singing module, legal module, calculation module, learning English module, weather overview module, news module, stock question and answer module, and recipe module Wait and ask A database of answers. The category database also uses vectors to store related question and answer sets. The voice recognition circuit 120 compares the voice vector with the question and answer set vector in the category database, finds the category database with a smaller gap between the two, and recognizes which category database the voice data may belong to. For example, the "buy stocks" paragraph has a small gap with the data in the stock Q&A module, the encyclopedia module, and the legal module. Therefore, the semantic meaning of the voice data is determined to be related to the stock Q&A module, the encyclopedia module, and the legal module. Answers are more likely to appear in the question and answer collections stored in the stock question and answer module, encyclopedia module, and legal module.

In step S220, the range of possible answers is first narrowed down to the stock Q&A module, the encyclopedia module, and the legal module. In step S230, the processing circuit 130 compares the voice vector with the question and answer set vectors in the stock question answering module, encyclopedia module, and legal module, and calculates the gap between the voice vector and the question answer set vector.

In step S240, the processing circuit 130 sequentially arranges the answers with the smallest gap to the largest, and the processing circuit 130 can set the number of ranked answers, for example, set the top 20 answers with the smallest gap. In step S250, the answer with the smallest gap is recommended to the user according to the ranking result. For example, the processing circuit 130 compares the question with the smallest difference in the stock question and answer module as "cannot buy stocks at the age of 18", so it will answer the user "No."

In one embodiment, the intelligent recommendation system 100 includes a historical answer database (not shown). The historical answer database is used to store questions or words previously asked by the user and answers recommended by the intelligent recommendation system 100. After the processing circuit 130 in the intelligent recommendation system 100 sorts the searched answers in step S240, it analyzes the historical recommended answers and the current recommended responses. The order of the answers to recommend to users to recommend answers. For example, a user asks "Which restaurant A and B are more delicious?" After the intelligent recommendation system 100 sorts the answers related to "good food", restaurant A is ranked at the forefront, and the number of previous questions the user has asked is Questions about restaurant A are higher than those about restaurant B. Therefore, the intelligent recommendation system 100 determines that the user is more interested in restaurant A and recommends restaurant A to the user first.

In one embodiment, the user asks "Which of the two restaurants A and B is more delicious?" After the intelligent recommendation system 100 sorts the answers related to "Delicious", restaurant A is ranked before restaurant B. The number of previous questions asked by the user is that the question about restaurant B is higher than that about restaurant A, so the intelligent recommendation system 100 determines that the user is more interested in restaurant B, and recommends restaurant B to the user first.

In an embodiment, the historical answer database may be an external database or built into the intelligent recommendation system 100. The historical answer database stores the interactive records between the intelligent recommendation system 100 and the user. The user can give feedback based on the answers recommended by the intelligent recommendation system 100, such as accuracy reports. The intelligent recommendation system 100 uses the historical answer database Record, make the recommended answer closer to the answer the user needs.

To sum up, the voice message is segmented by the voice segmentation circuit. The voice recognition circuit recognizes the semantic meaning of the voice data, and searches for multiple answers related to the semantic meaning of the voice data in a plurality of category databases. Provide relevant answers and calculate the gap, and recommend the answer with the smallest gap to the user to intelligently solve the user’s problem.

100‧‧‧Smart Recommendation System

110‧‧‧Speech segmentation circuit

120‧‧‧Voice recognition circuit

130‧‧‧Processing circuit

140-1, 140-2, 140-n‧‧‧Category database

Claims (8)

An intelligent recommendation system includes: a voice segmentation circuit for receiving a voice data and segmenting the voice data according to the content of the voice data; a voice recognition circuit for recognizing the voice based on the segmented voice data The semantics of the data; a plurality of category databases, each of which is used to store a plurality of answers; and a processing circuit for searching the semantics of the voice data in the category databases according to the semantics of the voice data The answers related to connectivity, wherein the processing circuit ranks the answers according to the semantic meaning of the speech data and the relevance between the answers, and recommends a recommended answer among the answers after the ranking, wherein the The voice recognition circuit is further used to convert the segmented voice data into a voice vector, and compare the voice vector with the answers in the category databases to determine that the voice data corresponds to the categories Data from at least one of the databases. The intelligent recommendation system according to claim 1, wherein the processing circuit is further used for calculating the gap between the voice vector and the answers of at least one of the categories of databases, and searching for the corresponding voice data according to the calculation result Some answers. The intelligent recommendation system according to claim 2, wherein the processing circuit further sorts the answers according to the calculation results, and arranges the answers in order from small gap to large, and the processing circuit further recommends the gaps in the answers The smallest result. The intelligent recommendation system according to claim 1, wherein the processing circuit sorts the answers according to the semantic meaning of the voice data and the relevance between the answers to obtain a ranking result, and according to the ranking result and a history A plurality of historical recommended answers stored in the answer database is recommended to recommend the recommended answer among the answers or one of the historical recommended answers. An intelligent recommendation method includes: receiving a voice data through a voice segmentation circuit, and segmenting the voice data according to the content of the voice data; through a voice recognition circuit, recognizing the semantic meaning of the voice data based on the segmented voice data ; According to the semantic meaning of the voice data, search for a plurality of answers related to the semantic meaning of the voice data in a plurality of category databases; analyze the relevance between the semantic meaning of the voice data and the answers, and then these answers Ranking; and recommending a recommended answer of the answers, wherein the step of identifying the semantic meaning of the voice data according to the segmented voice data includes: Convert the segmented voice data into a voice vector, and compare the voice vector with the answers in the category databases to determine whether the voice data corresponds to at least one of the category databases data of. The intelligent recommendation method of claim 5, wherein the step of searching for a plurality of answers related to the semantic meaning of the voice data in the category databases includes: the voice vector and at least one of the category databases Calculate the gap between the answers of the person and search for the answers corresponding to the voice data according to the calculation result. According to the intelligent recommendation method of claim 6, wherein the step of sorting the answers includes: sorting the answers according to the calculation result, and arranging the answers from the smallest gap to the largest one, wherein the processing circuit is further used for The result with the smallest gap among the answers is recommended. The intelligent recommendation method according to claim 5 further includes: sorting the answers according to the semantic meaning of the voice data and the relevance between the answers to obtain a sorting result, and according to the sorting result and passing through a history A plurality of historical recommended answers stored in the answer database is recommended to recommend the recommended answer among the answers or one of the historical recommended answers.

Images