WO2020240996A1 - Information processing device, information processing method, and program - Google Patents

Information processing device, information processing method, and program Download PDF

Info

Publication number
WO2020240996A1
WO2020240996A1 PCT/JP2020/011608 JP2020011608W WO2020240996A1 WO 2020240996 A1 WO2020240996 A1 WO 2020240996A1 JP 2020011608 W JP2020011608 W JP 2020011608W WO 2020240996 A1 WO2020240996 A1 WO 2020240996A1
Authority
WO
WIPO (PCT)
Prior art keywords
user
request
information processing
content
control unit
Prior art date
Application number
PCT/JP2020/011608
Other languages
French (fr)
Japanese (ja)
Inventor
昭彦 泉
賢司 久永
研二 小川
太一 下屋鋪
智哉 藤田
Original Assignee
ソニー株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by ソニー株式会社 filed Critical ソニー株式会社
Publication of WO2020240996A1 publication Critical patent/WO2020240996A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/903—Querying
    • G06F16/9035—Filtering based on additional data, e.g. user or group profiles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/907—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/908—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content

Definitions

  • This technology is related to information processing equipment. More specifically, the present invention relates to an information processing apparatus that processes a user's request based on an object of interest of the user, a processing method thereof, and a program that causes a computer to execute the processing method.
  • the user's interest is extracted from the document browsing history, action history, and conversation history.
  • a recognition result in which a preset keyword is voice-recognized is used. That is, it is necessary to set keywords for extracting the user's interest in advance.
  • the suitable keyword differs depending on the request from the user. Also, assuming a large number of purposes and extracting user interests from various perspectives, the number of combinations is enormous, and the scale and cost of hardware resources for extraction processing and result retention are not realistic. ..
  • This technology was created in view of such a situation, and in an information processing device that processes a user's request, the user's request is made based on the object of the user's interest extracted in response to the user's request.
  • the purpose is to process.
  • the present technology has been made to solve the above-mentioned problems, and the first aspect thereof is that after acquiring the user's request, the above-mentioned user's behavior history corresponding to the characteristics of the above-mentioned request is obtained.
  • An information processing device including a control unit that extracts an object of interest and processes the request based on the extracted object of interest, a processing method thereof, and a program. This has the effect of processing the request based on the object of interest of the user extracted from the user's behavior history corresponding to the characteristics of the user's request.
  • control unit may abstract the requirement to generate the features of the requirement. This has the effect of accurately searching the user's behavior history corresponding to the characteristics of the user's request.
  • control unit may use a neural network to abstract the above requirements.
  • control unit may acquire an exclusion phrase to be excluded when generating the feature of the request and exclude it from the action history of the user. This has the effect of obtaining candidates that become noise when searching the user's behavior history.
  • control unit may acquire the exclusion phrase by using a neural network.
  • control unit may extract the target of interest of the user based on the score of the metadata attached to the behavior history of the user. This has the effect of extracting the object of interest of the user based on the quantified information.
  • control unit may generate the score of the metadata based on the degree of matching with the feature of the request, the number of times in the action history, and the novelty. This has the effect of extracting the object of interest of the user based on these specific information. Further, in this case, the control unit may extract the value of the metadata having the higher score as the object of interest of the user.
  • control unit registers the action content of the user as the action history of the user in a predetermined database, obtains the request of the user, and then acquires the action history of the user from the database. You may try to search. This has the effect of searching the user's behavior history based on the information stored in the database.
  • the behavior history of the user is, for example, the viewing history of the content by the user. That is, it has the effect of processing the request based on the object of interest of the user extracted from the viewing history of the content by the user.
  • FIG. 1 is a diagram showing an overall configuration example of an information processing apparatus according to an embodiment of the present technology.
  • an information processing device reproduces content in response to a viewing request from a user and performs processing in response to a subsequent processing request from the user based on the playback history of the content.
  • the user who makes the viewing request and the user who makes the subsequent processing request are the same person.
  • Such an information processing device that performs processing in response to a processing request from a user has been put into practical use as, for example, an AI (Artificial Intelligence) agent, a smart speaker, or the like. That is, the information processing device according to the embodiment of the present technology can broadly include, for example, a television receiver, an agent terminal, a server device, and the like.
  • This information processing device includes a user viewing request receiving unit 101, a content viewing time processing unit 102, a viewing behavior database 300, a user request processing unit 104, and a user processing request receiving unit 140.
  • the user viewing request receiving unit 101 receives a content viewing request from the user.
  • the user viewing request receiving unit 101 is realized by, for example, a remote control light receiving unit of a content reproduction device.
  • the content viewing time processing unit 102 performs processing at the time of content viewing by the user.
  • the content viewing time processing unit 102 includes a viewing behavior registration unit 120.
  • the viewing behavior registration unit 120 documents the content to be viewed by the user, extracts the metadata of the content, and registers the documented content in the viewing behavior database 300 together with the viewing date and time.
  • the viewing behavior database 300 is a database that stores the content of the content to be viewed by the user documented by the content viewing processing unit 102 and the metadata of the content together with the viewing date and time. That is, the viewing behavior database 300 stores the user's content viewing history. The stored contents of the viewing behavior database 300 are searched by the user request processing unit 104.
  • the user processing request receiving unit 140 receives a processing request from the user.
  • the user processing request receiving unit 140 instructs the user request processing unit 104 to search the viewing behavior database 300 for the user's request.
  • the user request processing unit 104 searches the viewing behavior database 300 for the user's request according to the instruction from the user processing request receiving unit 140, extracts the user's preference (object of interest), and based on the preference. It handles requests from users.
  • the user request processing unit 104 includes a preference extraction unit 150 and a user request processing execution unit 160.
  • the preference extraction unit 150 searches the viewing behavior database 300 for the user's request and extracts the preference.
  • the preference extraction unit 150 abstracts the user's request and searches the viewing behavior database 300 according to the characteristics of the abstracted request. At that time, as will be described later, inappropriate ones may be excluded.
  • the preference extraction unit 150 obtains the score of the metadata associated with each content from the higher-ranked results in the obtained search results, and uses the value of the metadata with the higher score as the preference.
  • the user request processing execution unit 160 processes a request from a user based on the preferences extracted by the preference extraction unit 150.
  • the user request processing execution unit 160 gives priority to the information having a high degree of applicability to the preference and presents the information to the user.
  • the user viewing request receiving unit 101, the content viewing time processing unit 102, the user request processing unit 104, and the user processing request receiving unit 140 are examples of the control units described in the claims.
  • This control unit can be realized by, for example, a central processing unit (CPU: Central Processing Unit) of a computer.
  • the information processing apparatus searches the content viewing history of the user corresponding to the characteristics of the request after receiving the processing request from the user, extracts the user's preference, and extracts the user's preference.
  • by extracting the user's preferences after accepting the user's request it is possible to make an accurate proposal.
  • FIG. 2 is a diagram showing an example of processing contents at the time of content viewing in the embodiment of the present technology. Further, FIG. 3 is a flow chart showing an example of a processing procedure at the time of content viewing in the embodiment of the present technology.
  • the user's preference is not extracted when viewing the content.
  • the content content is registered in the viewing behavior database 300.
  • the viewing behavior registration unit 120 documents the content of the viewing content 200 as the content content 121 (step S911). At the time of this documenting, it is ideal that the content of the viewing content 200 is analyzed by a computer and the synopsis is automatically generated. On the other hand, as a simple realization method, it is more realistic to use the information registered in the TV program information or the like as a synopsis of the viewing content 200.
  • the viewing behavior registration unit 120 extracts metadata 122 from the viewing content 200 (step S912).
  • the metadata given to the contents by broadcasting stations including Internet broadcasting.
  • the nouns included in the content of the viewing content 200 documented in step S911 may be extracted, classified into categories, and the metadata 122 may be collected.
  • the synopsis is automatically generated in step S911, the metadata 122 is not extracted at this point, but the metadata 122 is extracted from the documented content content 121 when responding to the user's processing request. It may be.
  • the viewing behavior registration unit 120 registers information about the content viewed by the user in the viewing behavior database 300 (step S913). At this time, the viewing date and time and the documented content content 121 are registered in the viewing behavior database 300. The method of registration in the viewing behavior database 300 is the same as that of a known search engine. Corresponding metadata 122 is associated with the documented content content 121.
  • FIG. 4 is a diagram showing an example of processing contents at the time of user request in the embodiment of the present technology. Further, FIG. 5 is a flow chart showing an example of a processing procedure at the time of user request in the embodiment of the present technology.
  • the preference based on the user's past content viewing behavior is extracted after the user's request is known.
  • the user processing request receiving unit 140 receives a processing request from the user (step S921).
  • the user processing request receiving unit 140 acquires, for example, a request such as "I want to go to a restaurant" from a user.
  • the preference extraction unit 150 abstracts the request from the user (step S922). For this abstraction, it is conceivable to use a neural network as described later.
  • the preference extraction unit 150 obtains an abstract request of "noun” cooking “+ verb” eating “” in response to a request from a user such as "want to go to a restaurant”.
  • the preference extraction unit 150 searches the viewing behavior database 300 for the abstracted request obtained in step S922 (step S923).
  • a search result for example, the following information can be obtained.
  • --Address in viewing behavior database 300 of content content 131 --Latest viewing date and time --Number of views --Matching score (likelihood) in search
  • the matching score the content that matches the search term better shows a higher score.
  • the preference extraction unit 150 generates an excluded phrase to be excluded when generating the feature of the request (step S924). To generate this exclusion phrase, it is conceivable to use a neural network as described later.
  • the preference extraction unit 150 obtains the exclusion phrase "noun" meal “+ verb” make "” in response to a request from the user, for example, "I want to go to a restaurant”.
  • the preference extraction unit 150 searches the viewing behavior database 300 for the exclusion phrase obtained in step S924 (step S925).
  • a search result for example, the following information can be obtained. --Address in the viewing behavior database 300 of the content content 131
  • the preference extraction unit 150 excludes the search results obtained in step S925 from the search results obtained in step S925 to obtain the entire search results (step S926).
  • FIG. 6 is a diagram showing an example of excluding the search result of the exclusion phrase in the embodiment of the present technology.
  • a in the figure is an example of the search result obtained in step S923.
  • search results such as content AAAA, content BBBB, and content CCCC are shown for the user's request.
  • a in the figure is an example of the search result obtained in step S925.
  • search results such as content ZZZZ and content BBBB are shown for the exclusion phrase.
  • the content BBBB is excluded because it is included in the search result of the exclusion phrase.
  • the result excluding the content BBBB is obtained as the overall search result.
  • the preference extraction unit 150 aggregates the scores of each metadata 132 associated with the content content 131 from the higher-ranked results in the overall search results (step S927).
  • the preference extraction unit 150 obtains each correction score for the content content 131 that is higher in the overall search results.
  • the "score” is the matching score (likelihood) in the above-mentioned search.
  • the "newness” is a value that is larger as the viewing date and time is more recent, and is a value from 0 to 1.
  • the metadata score associated with each content shall be the correction score of the associated content.
  • the total score is used. For example, metadata contained in content that is repeatedly viewed, such as a serial drama, has a high score. Also, if you are watching various contents that are not continuous but contain the same metadata, you will get a high score.
  • FIG. 7 is a diagram showing an example of calculating a metadata score in an embodiment of the present technology.
  • the correction score is calculated from the above formula.
  • the content AAAA correction score 72
  • the content CCCC correction score 126
  • the score of the metadata "Food: French cuisine” is calculated. Since the metadata of content AAAA and content DDDD includes “food: French cuisine”, their respective correction scores are added. On the other hand, since the content CCCC metadata does not include “food: French cuisine”, the value to be added is 0. Therefore, the total score of the metadata "Food: French cuisine” is calculated as 72 + 0 + 120 + ...
  • the preference extraction unit 150 uses the value of the metadata having the higher score calculated as described above as the preference (step S928).
  • the preference for example, metadata values such as "French cuisine”, “Michelin”, and "ramen with a queue” are extracted as preferences.
  • the user request processing execution unit 160 processes the request from the user based on the preference extracted by the preference extraction unit 150 (step S929). For example, a list of restaurants is acquired, and stores with a high degree of appetite are presented to the user as promising candidates. In the above example, the user is presented with a restaurant corresponding to any of "French cuisine”, “Michelin”, and "Ramen with a line”.
  • FIG. 8 is a diagram showing a configuration example for receiving a user processing request in the embodiment of the present technology.
  • a trained neural network is used to convert the content requested by the user into an abstract expression or an exclusion phrase.
  • DNN Deep Neural Network
  • the user processing request receiving unit 140 includes a voice recognizer 610.
  • the voice recognizer 610 voice recognizes the user-requested voice and, as a result, produces a text output.
  • the text user request by the voice recognizer 610 is supplied to the DNN 620.
  • DNN620 includes DNN621 for abstract expression and DNN622 for exclusion phrase.
  • DNN621 and 622 are both trained DNNs. These DNNs 621 and 622 supply the user-requested abstract expressions and exclusion phrases to the viewing behavior database 300 for use in the search.
  • FIG. 9 is a diagram showing a configuration example when learning DNN620 in the embodiment of the present technology.
  • the user's content viewing history corresponding to the feature of the request is searched and the user's preference is extracted.
  • User requests can be processed based on the reference.
  • the content (synopsis) of the book is registered in the database and linked with the metadata. For example, suppose the book was "The Age of Mozart and Beethoven.” After that, when the user requests "play BGM”, the database is searched using "song", “kake”, etc. as search terms. As a result of the search, “Mozart” and “Beethoven” are searched. As a result, Mozart and Beethoven songs are provided to the user. That is, it is possible to extract a preference for playing music from an action history that is different from listening to music.
  • location information such as sightseeing spots and commercial areas is specified from the GPS information of the terminal carried by the user, and the user's behavior is documented based on the specified location information and registered in the database, and the preference is set. It may be extracted after the fact.
  • DNN In the above-described embodiment, it is assumed that DNN is used to convert the content requested by the user into an abstract expression or an exclusion phrase, but these conversion processes are realized by a technique other than DNN. You may.
  • Exclusion phrase In the above-described embodiment, an example of improving the search accuracy by using the exclusion phrase has been described. However, if the accuracy of searching the user request is improved by improving the search technique, the exclusion phrase may not be used. Good. That is, if the noise of the search result is reduced, the exclusion phrase does not necessarily have to be used.
  • the processing procedure described in the above-described embodiment may be regarded as a method having these series of procedures, and as a program for causing a computer to execute these series of procedures or as a recording medium for storing the program. You may catch it.
  • a recording medium for example, a CD (Compact Disc), MD (MiniDisc), DVD (Digital Versatile Disc), memory card, Blu-ray Disc (Blu-ray (registered trademark) Disc) and the like can be used.
  • the present technology can have the following configurations.
  • Control that after acquiring a user's request, an object of interest of the user is extracted from the behavior history of the user corresponding to the feature of the request, and the request is processed based on the extracted object of interest.
  • An information processing device including a unit.
  • (4) The information according to any one of (1) to (3) above, in which the control unit acquires an exclusion phrase to be excluded when generating the feature of the request and excludes it from the behavior history of the user. Processing equipment.
  • control unit acquires the exclusion phrase using a neural network.
  • control unit extracts an object of interest of the user based on a score of metadata associated with the behavior history of the user.
  • control unit generates a score of the metadata based on the degree of matching with the feature of the request, the number of times in the action history, and the novelty.
  • control unit extracts the value of the metadata having the higher score as the object of interest of the user.
  • the control unit registers the action content of the user as the action history of the user in a predetermined database, acquires the request of the user, and then searches the action history of the user from the database (1).
  • the information processing apparatus according to any one of (8) to (8).
  • the information processing device according to any one of (1) to (9) above, wherein the action history of the user is a viewing history of the content by the user.
  • (11) The procedure for the control unit to acquire the user's request and A procedure in which the control unit extracts an object of interest of the user from the behavior history of the user corresponding to the feature of the request.
  • An information processing method in which the control unit includes a procedure for processing the request based on the extracted object of interest.

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Library & Information Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

Provided is an information processing device for processing a request from a user, wherein the request from the user is processed on the basis of a user preference extracted in accordance with the user's request.　After a request from a user is acquired, an object of the user's preference is extracted from a user behavior history corresponding to a feature of the user's request. The user's request is then processed on the basis of the extracted object of the user's preference. It is assumed that the user's behavior history has been registered in a database or the like prior to the acquisition of the user's request.

Description

情報処理装置、情報処理方法、および、プログラムInformation processing equipment, information processing methods, and programs
　本技術は、情報処理装置に関する。詳しくは、ユーザの興味の対象に基づいてユーザの要求を処理する情報処理装置、および、その処理方法ならびに当該方法をコンピュータに実行させるプログラムに関する。 This technology is related to information processing equipment. More specifically, the present invention relates to an information processing apparatus that processes a user's request based on an object of interest of the user, a processing method thereof, and a program that causes a computer to execute the processing method.
　従来、ユーザの要求に対してそのユーザの興味の傾向を判断して情報提示を行う技術が知られている。例えば、ユーザに提示する候補の文書に対し、行動履歴や会話内容なども考慮して、ユーザの興味の傾向を求め、ユーザに提示するか否かを判断する情報提示システムが提案されている（例えば、特許文献１参照。）。 Conventionally, there is known a technique for presenting information by judging a tendency of interest of the user in response to a request of the user. For example, an information presentation system has been proposed in which a candidate document to be presented to a user is asked for a tendency of interest of the user in consideration of behavior history, conversation content, etc., and whether or not to present the document to the user is determined. For example, see Patent Document 1.).
特許第４１１４５０９号公報Japanese Patent No. 4114509
　上述の従来技術では、文書閲覧履歴、行動履歴、会話履歴からユーザの興味を抽出している。この従来技術では、会話履歴からユーザの興味を抽出する際に、予め設定されたキーワードが音声認識された認識結果を用いている。すなわち、ユーザの興味を抽出するためのキーワードは予め設定しておく必要がある。しかしながら、そのようなキーワードを適切に設定するためには、本来の目的を明確にしておく必要があり、ユーザからの要求によって適するキーワードは異なると考えられる。また、多数の目的を想定して、色々な観点でユーザの興味を抽出しておくのは、組合せが膨大となり、抽出処理や結果保持のためのハードウェアリソースの規模やコストが現実的ではない。 In the above-mentioned conventional technique, the user's interest is extracted from the document browsing history, action history, and conversation history. In this conventional technique, when extracting a user's interest from a conversation history, a recognition result in which a preset keyword is voice-recognized is used. That is, it is necessary to set keywords for extracting the user's interest in advance. However, in order to properly set such a keyword, it is necessary to clarify the original purpose, and it is considered that the suitable keyword differs depending on the request from the user. Also, assuming a large number of purposes and extracting user interests from various perspectives, the number of combinations is enormous, and the scale and cost of hardware resources for extraction processing and result retention are not realistic. ..
　本技術はこのような状況に鑑みて生み出されたものであり、ユーザの要求を処理する情報処理装置において、ユーザの要求に応じて抽出されたユーザの興味の対象に基づいて、ユーザの要求を処理することを目的とする。 This technology was created in view of such a situation, and in an information processing device that processes a user's request, the user's request is made based on the object of the user's interest extracted in response to the user's request. The purpose is to process.
　本技術は、上述の問題点を解消するためになされたものであり、その第１の側面は、ユーザの要求を取得した後に、上記要求の特徴に対応する上記ユーザの行動履歴から上記ユーザの興味の対象を抽出し、その抽出した上記興味の対象に基づいて上記要求を処理する制御部を具備する情報処理装置、その処理方法、および、プログラムである。これにより、ユーザの要求の特徴に対応するユーザの行動履歴から抽出されたユーザの興味の対象に基づいて、要求を処理するという作用をもたらす。 The present technology has been made to solve the above-mentioned problems, and the first aspect thereof is that after acquiring the user's request, the above-mentioned user's behavior history corresponding to the characteristics of the above-mentioned request is obtained. An information processing device including a control unit that extracts an object of interest and processes the request based on the extracted object of interest, a processing method thereof, and a program. This has the effect of processing the request based on the object of interest of the user extracted from the user's behavior history corresponding to the characteristics of the user's request.
　また、この第１の側面において、上記制御部は、上記要求を抽象化して上記要求の特徴を生成するようにしてもよい。これにより、ユーザの要求の特徴に対応するユーザの行動履歴を的確に検索するという作用をもたらす。この場合において、上記制御部は、ニューラルネットワークを用いて上記要求を抽象化するようにしてもよい。 Further, in this first aspect, the control unit may abstract the requirement to generate the features of the requirement. This has the effect of accurately searching the user's behavior history corresponding to the characteristics of the user's request. In this case, the control unit may use a neural network to abstract the above requirements.
　また、この第１の側面において、上記制御部は、上記要求の特徴を生成する際に除外すべき除外フレーズを取得して、上記ユーザの行動履歴から除外するようにしてもよい。これにより、ユーザの行動履歴を検索する際のノイズとなる候補を得るという作用をもたらす。この場合において、上記制御部は、ニューラルネットワークを用いて上記除外フレーズを取得するようにしてもよい。 Further, in this first aspect, the control unit may acquire an exclusion phrase to be excluded when generating the feature of the request and exclude it from the action history of the user. This has the effect of obtaining candidates that become noise when searching the user's behavior history. In this case, the control unit may acquire the exclusion phrase by using a neural network.
　また、この第１の側面において、上記制御部は、上記ユーザの行動履歴に付随するメタデータのスコアに基づいて上記ユーザの興味の対象を抽出するようにしてもよい。これにより、数値化された情報に基づいてユーザの興味の対象を抽出するという作用をもたらす。この場合において、上記制御部は、上記要求の特徴とのマッチング度合い、上記行動履歴における回数および新しさに基づいて上記メタデータのスコアを生成するようにしてもよい。これにより、これらの具体的な情報に基づいてユーザの興味の対象を抽出するという作用をもたらす。また、この場合において、上記制御部は、上記スコアが上位となる上記メタデータの値を上記ユーザの興味の対象として抽出するようにしてもよい。 Further, in this first aspect, the control unit may extract the target of interest of the user based on the score of the metadata attached to the behavior history of the user. This has the effect of extracting the object of interest of the user based on the quantified information. In this case, the control unit may generate the score of the metadata based on the degree of matching with the feature of the request, the number of times in the action history, and the novelty. This has the effect of extracting the object of interest of the user based on these specific information. Further, in this case, the control unit may extract the value of the metadata having the higher score as the object of interest of the user.
　また、この第１の側面において、上記制御部は、上記ユーザの行動内容を上記ユーザの行動履歴として所定のデータベースに登録し、上記ユーザの要求を取得した後に上記データベースから上記ユーザの行動履歴を検索するようにしてもよい。これにより、データベースに蓄積した情報に基づいてユーザの行動履歴を検索するという作用をもたらす。 Further, in the first aspect, the control unit registers the action content of the user as the action history of the user in a predetermined database, obtains the request of the user, and then acquires the action history of the user from the database. You may try to search. This has the effect of searching the user's behavior history based on the information stored in the database.
　また、この第１の側面において、上記ユーザの行動履歴は、例えば、上記ユーザによるコンテンツの視聴履歴である。すなわち、ユーザによるコンテンツの視聴履歴から抽出されたユーザの興味の対象に基づいて、要求を処理するという作用をもたらす。 Further, in this first aspect, the behavior history of the user is, for example, the viewing history of the content by the user. That is, it has the effect of processing the request based on the object of interest of the user extracted from the viewing history of the content by the user.
本技術の実施の形態における情報処理装置の全体構成例を示す図である。It is a figure which shows the whole structure example of the information processing apparatus in embodiment of this technique. 本技術の実施の形態におけるコンテンツ視聴時の処理内容の一例を示す図である。It is a figure which shows an example of the processing content at the time of content viewing in embodiment of this technology. 本技術の実施の形態におけるコンテンツ視聴時の処理手順の一例を示す流れ図である。It is a flow chart which shows an example of the processing procedure at the time of content viewing in embodiment of this technology. 本技術の実施の形態におけるユーザ要求時の処理内容の一例を示す図である。It is a figure which shows an example of the processing content at the time of a user request in embodiment of this technology. 本技術の実施の形態におけるユーザ要求時の処理手順の一例を示す流れ図である。It is a flow chart which shows an example of the processing procedure at the time of a user request in embodiment of this technique. 本技術の実施の形態において除外フレーズの検索結果を除外する例を示す図である。It is a figure which shows the example which excludes the search result of the exclusion phrase in embodiment of this technique. 本技術の実施の形態においてメタデータのスコアを算出する例を示す図である。It is a figure which shows the example of calculating the score of metadata in embodiment of this technique. 本技術の実施の形態においてユーザ処理要求を受け付ける構成例を示す図である。It is a figure which shows the configuration example which accepts the user processing request in embodiment of this technique. 本技術の実施の形態においてＤＮＮ６２０を学習させる際の構成例を示す図である。It is a figure which shows the configuration example at the time of learning DNN620 in the embodiment of this technique.
　以下、本技術を実施するための形態（以下、実施の形態と称する）について説明する。説明は以下の順序により行う。
　１．実施の形態
　２．変形例 Hereinafter, embodiments for carrying out the present technology (hereinafter referred to as embodiments) will be described. The explanation will be given in the following order.
1. 1. Embodiment 2. Modification example
　＜１．実施の形態＞
　［全体構成］
　図１は、本技術の実施の形態における情報処理装置の全体構成例を示す図である。
 ここでは、ユーザからの視聴要求に応じてコンテンツを再生し、コンテンツの再生履歴に基づいてその後のユーザからの処理要求に応じて処理を行う情報処理装置を想定する。ただし、視聴要求を行うユーザと、その後の処理要求を行うユーザは同一人物であることを想定する。このようなユーザからの処理要求に応じて処理を行う情報処理装置は、例えば、ＡＩ（Artificial Intelligence）エージェントやスマートスピーカー等として実用化されている。すなわち、本技術の実施の形態における情報処理装置は、例えば、テレビ受信機、エージェント端末、サーバ装置などを広く含み得る。 <1. Embodiment>
[overall structure]
FIG. 1 is a diagram showing an overall configuration example of an information processing apparatus according to an embodiment of the present technology.
Here, it is assumed that an information processing device reproduces content in response to a viewing request from a user and performs processing in response to a subsequent processing request from the user based on the playback history of the content. However, it is assumed that the user who makes the viewing request and the user who makes the subsequent processing request are the same person. Such an information processing device that performs processing in response to a processing request from a user has been put into practical use as, for example, an AI (Artificial Intelligence) agent, a smart speaker, or the like. That is, the information processing device according to the embodiment of the present technology can broadly include, for example, a television receiver, an agent terminal, a server device, and the like.
　この情報処理装置は、ユーザ視聴要求受付部１０１と、コンテンツ視聴時処理部１０２と、視聴行動データベース３００と、ユーザ要求時処理部１０４と、ユーザ処理要求受付部１４０とを備える。 This information processing device includes a user viewing request receiving unit 101, a content viewing time processing unit 102, a viewing behavior database 300, a user request processing unit 104, and a user processing request receiving unit 140.
　ユーザ視聴要求受付部１０１は、ユーザからのコンテンツ視聴要求を受け付けるものである。このユーザ視聴要求受付部１０１は、例えば、コンテンツ再生装置のリモコン受光部などにより実現される。 The user viewing request receiving unit 101 receives a content viewing request from the user. The user viewing request receiving unit 101 is realized by, for example, a remote control light receiving unit of a content reproduction device.
　コンテンツ視聴時処理部１０２は、ユーザによるコンテンツ視聴時の処理を行うものである。このコンテンツ視聴時処理部１０２は、視聴行動登録部１２０を備える。この視聴行動登録部１２０は、ユーザが視聴するコンテンツの内容を文書化するとともに、そのコンテンツのメタデータを抽出して、文書化されたコンテンツを視聴日時とともに視聴行動データベース３００に登録する。 The content viewing time processing unit 102 performs processing at the time of content viewing by the user. The content viewing time processing unit 102 includes a viewing behavior registration unit 120. The viewing behavior registration unit 120 documents the content to be viewed by the user, extracts the metadata of the content, and registers the documented content in the viewing behavior database 300 together with the viewing date and time.
　視聴行動データベース３００は、コンテンツ視聴時処理部１０２によって文書化されたユーザが視聴するコンテンツの内容と、そのコンテンツのメタデータとを、視聴日時とともに記憶するデータベースである。すなわち、この視聴行動データベース３００は、ユーザのコンテンツ視聴履歴を記憶する。この視聴行動データベース３００の記憶内容は、ユーザ要求時処理部１０４によって検索される。 The viewing behavior database 300 is a database that stores the content of the content to be viewed by the user documented by the content viewing processing unit 102 and the metadata of the content together with the viewing date and time. That is, the viewing behavior database 300 stores the user's content viewing history. The stored contents of the viewing behavior database 300 are searched by the user request processing unit 104.
　ユーザ処理要求受付部１４０は、ユーザからの処理要求を受け付けるものである。このユーザ処理要求受付部１４０は、ユーザからの処理要求を受け付けると、ユーザの要求について視聴行動データベース３００を検索するようユーザ要求時処理部１０４に指示する。 The user processing request receiving unit 140 receives a processing request from the user. When the user processing request receiving unit 140 receives the processing request from the user, the user processing request receiving unit 140 instructs the user request processing unit 104 to search the viewing behavior database 300 for the user's request.
　ユーザ要求時処理部１０４は、ユーザ処理要求受付部１４０からの指示に従って、ユーザの要求について視聴行動データベース３００を検索し、ユーザのプリファレンス（興味の対象）を抽出して、そのプリファレンスに基づいてユーザからの要求を処理するものである。このユーザ要求時処理部１０４は、プリファレンス抽出部１５０と、ユーザ要求処理実行部１６０とを備える。 The user request processing unit 104 searches the viewing behavior database 300 for the user's request according to the instruction from the user processing request receiving unit 140, extracts the user's preference (object of interest), and based on the preference. It handles requests from users. The user request processing unit 104 includes a preference extraction unit 150 and a user request processing execution unit 160.
　プリファレンス抽出部１５０は、ユーザの要求について視聴行動データベース３００を検索して、プリファレンスを抽出するものである。このプリファレンス抽出部１５０は、ユーザの要求を抽象化して、その抽象化された要求の特徴によって視聴行動データベース３００を検索する。その際、後述するように、不適切なものを除外するようにしてもよい。このプリファレンス抽出部１５０は、得られた検索結果の中で上位の結果から、各コンテンツに紐づいているメタデータのスコアを求め、スコアが上位のメタデータの値をプリファレンスとする。 The preference extraction unit 150 searches the viewing behavior database 300 for the user's request and extracts the preference. The preference extraction unit 150 abstracts the user's request and searches the viewing behavior database 300 according to the characteristics of the abstracted request. At that time, as will be described later, inappropriate ones may be excluded. The preference extraction unit 150 obtains the score of the metadata associated with each content from the higher-ranked results in the obtained search results, and uses the value of the metadata with the higher score as the preference.
　ユーザ要求処理実行部１６０は、プリファレンス抽出部１５０によって抽出されたプリファレンスに基づいてユーザからの要求を処理するものである。このユーザ要求処理実行部１６０は、例えば、情報を提示する際に、プリファレンスへの該当度合いが高いものを優先して、ユーザに提示する。 The user request processing execution unit 160 processes a request from a user based on the preferences extracted by the preference extraction unit 150. When presenting information, for example, the user request processing execution unit 160 gives priority to the information having a high degree of applicability to the preference and presents the information to the user.
　なお、ユーザ視聴要求受付部１０１、コンテンツ視聴時処理部１０２、ユーザ要求時処理部１０４、および、ユーザ処理要求受付部１４０は、特許請求の範囲に記載の制御部の一例である。この制御部は、例えば、コンピュータの中央処理装置（ＣＰＵ：Central Processing Unit）などにより実現され得る。 The user viewing request receiving unit 101, the content viewing time processing unit 102, the user request processing unit 104, and the user processing request receiving unit 140 are examples of the control units described in the claims. This control unit can be realized by, for example, a central processing unit (CPU: Central Processing Unit) of a computer.
　このように、この実施の形態における情報処理装置は、ユーザから処理要求を受け付けた後に、その要求の特徴に対応するユーザのコンテンツ視聴履歴を検索して、ユーザのプリファレンスを抽出し、抽出したプリファレンスに基づいてユーザの要求を処理する。これにより、例えば、サッカーの試合を視聴した人が、その後、「今日暇なので何かをしたい」と言った際に、サッカーのチケット購入を提案することができる。ユーザの要求としてどのようなものがあるかは、可能性が膨大であり、事前にその回答を用意しておくことは現実的ではない。この点、ユーザの要求を受け付けた後にユーザのプリファレンスを抽出することにより、的確な提案を行うことが可能となる。 As described above, the information processing apparatus according to this embodiment searches the content viewing history of the user corresponding to the characteristics of the request after receiving the processing request from the user, extracts the user's preference, and extracts the user's preference. Process user requests based on preferences. This allows, for example, a person who has watched a soccer match to propose to buy a soccer ticket when he or she later says, "I am free today and want to do something." There are enormous possibilities for what kind of user's request is, and it is not realistic to prepare the answer in advance. In this regard, by extracting the user's preferences after accepting the user's request, it is possible to make an accurate proposal.
　［コンテンツ視聴時処理］
　図２は、本技術の実施の形態におけるコンテンツ視聴時の処理内容の一例を示す図である。また、図３は、本技術の実施の形態におけるコンテンツ視聴時の処理手順の一例を示す流れ図である。 [Processing when viewing content]
FIG. 2 is a diagram showing an example of processing contents at the time of content viewing in the embodiment of the present technology. Further, FIG. 3 is a flow chart showing an example of a processing procedure at the time of content viewing in the embodiment of the present technology.
　この実施の形態では、コンテンツ視聴時には、ユーザのプリファレンスの抽出は行わない。その一方で、その後に高速にプリファレンスを抽出できるようにするために、コンテンツ内容を視聴行動データベース３００に登録する。 In this embodiment, the user's preference is not extracted when viewing the content. On the other hand, in order to enable high-speed extraction of preferences thereafter, the content content is registered in the viewing behavior database 300.
　視聴行動登録部１２０は、視聴コンテンツ２００の内容を、コンテンツ内容１２１として文書化する（ステップＳ９１１）。この文書化の際には、視聴コンテンツ２００の内容をコンピュータにより解析して、そのあらすじを自動生成することが理想的である。一方、簡易な実現手法としては、テレビ番組情報などに登録されている情報を視聴コンテンツ２００のあらすじとして利用することが、より現実的である。 The viewing behavior registration unit 120 documents the content of the viewing content 200 as the content content 121 (step S911). At the time of this documenting, it is ideal that the content of the viewing content 200 is analyzed by a computer and the synopsis is automatically generated. On the other hand, as a simple realization method, it is more realistic to use the information registered in the TV program information or the like as a synopsis of the viewing content 200.
　次に、視聴行動登録部１２０は、視聴コンテンツ２００からメタデータ１２２を抽出する（ステップＳ９１２）。簡易な実現手法としては、ネット放送を含む放送局がコンテンツに付与しているメタデータを利用することが考えられる。また、ステップＳ９１１において文書化された視聴コンテンツ２００の内容に含まれる名詞を抽出して、カテゴリに分類して、メタデータ１２２を収集してもよい。なお、ステップＳ９１１においてあらすじを自動生成した場合には、この時点でメタデータ１２２を抽出せずに、ユーザの処理要求への対応時に、文書化済のコンテンツ内容１２１からメタデータ１２２を抽出するようにしてもよい。 Next, the viewing behavior registration unit 120 extracts metadata 122 from the viewing content 200 (step S912). As a simple realization method, it is conceivable to use the metadata given to the contents by broadcasting stations including Internet broadcasting. Further, the nouns included in the content of the viewing content 200 documented in step S911 may be extracted, classified into categories, and the metadata 122 may be collected. When the synopsis is automatically generated in step S911, the metadata 122 is not extracted at this point, but the metadata 122 is extracted from the documented content content 121 when responding to the user's processing request. It may be.
　次に、視聴行動登録部１２０は、ユーザが視聴したコンテンツに関する情報を視聴行動データベース３００に登録する（ステップＳ９１３）。このとき、視聴行動データベース３００には、視聴日時と、文書化したコンテンツ内容１２１を登録する。視聴行動データベース３００における登録の手法は、公知の検索エンジンと同様である。文書化したコンテンツ内容１２１には、対応するメタデータ１２２が紐付けられる。 Next, the viewing behavior registration unit 120 registers information about the content viewed by the user in the viewing behavior database 300 (step S913). At this time, the viewing date and time and the documented content content 121 are registered in the viewing behavior database 300. The method of registration in the viewing behavior database 300 is the same as that of a known search engine. Corresponding metadata 122 is associated with the documented content content 121.
　［ユーザ要求時処理］
　図４は、本技術の実施の形態におけるユーザ要求時の処理内容の一例を示す図である。また、図５は、本技術の実施の形態におけるユーザ要求時の処理手順の一例を示す流れ図である。 [Processing at user request]
FIG. 4 is a diagram showing an example of processing contents at the time of user request in the embodiment of the present technology. Further, FIG. 5 is a flow chart showing an example of a processing procedure at the time of user request in the embodiment of the present technology.
　この実施の形態では、ユーザの過去のコンテンツ視聴行動に基づくプリファレンスの抽出を、ユーザからの要求が分かった後で行う。 In this embodiment, the preference based on the user's past content viewing behavior is extracted after the user's request is known.
　また、ユーザ処理要求受付部１４０は、ユーザからの処理の要求を受け付ける（ステップＳ９２１）。このユーザ処理要求受付部１４０は、例えば、ユーザから「レストランに行きたい」などの要求を取得する。 Further, the user processing request receiving unit 140 receives a processing request from the user (step S921). The user processing request receiving unit 140 acquires, for example, a request such as "I want to go to a restaurant" from a user.
　次に、プリファレンス抽出部１５０は、ユーザからの要求を抽象化する（ステップＳ９２２）。この抽象化のために、後述するように、ニューラルネットワークを利用することが考えられる。このプリファレンス抽出部１５０は、例えば、ユーザからの「レストランに行きたい」といった要求に対して、『名詞「料理」＋動詞「食べる」』という抽象化された要求を得る。 Next, the preference extraction unit 150 abstracts the request from the user (step S922). For this abstraction, it is conceivable to use a neural network as described later. The preference extraction unit 150 obtains an abstract request of "noun" cooking "+ verb" eating "" in response to a request from a user such as "want to go to a restaurant".
　次に、プリファレンス抽出部１５０は、ステップＳ９２２で得られた抽象化された要求について、視聴行動データベース３００を検索する（ステップＳ９２３）。検索結果として、例えば、以下の情報が得られる。
　　- コンテンツ内容１３１の視聴行動データベース３００内のアドレス
　　- 最新の視聴日時
　　- 視聴回数
　　- 検索におけるマッチングスコア（尤度）
なお、マッチングスコアは、検索語によくマッチしているコンテンツほど高いスコアを示す。 Next, the preference extraction unit 150 searches the viewing behavior database 300 for the abstracted request obtained in step S922 (step S923). As a search result, for example, the following information can be obtained.
--Address in viewing behavior database 300 of content content 131 --Latest viewing date and time --Number of views --Matching score (likelihood) in search
As for the matching score, the content that matches the search term better shows a higher score.
　次に、プリファレンス抽出部１５０は、要求の特徴を生成する際に除外すべき除外フレーズ（excluded phrase）を生成する（ステップＳ９２４）。この除外フレーズ生成のために、後述するように、ニューラルネットワークを利用することが考えられる。このプリファレンス抽出部１５０は、例えば、ユーザからの「レストランに行きたい」といった要求に対して、『名詞「食事」＋動詞「作る」』という除外フレーズを得る。 Next, the preference extraction unit 150 generates an excluded phrase to be excluded when generating the feature of the request (step S924). To generate this exclusion phrase, it is conceivable to use a neural network as described later. The preference extraction unit 150 obtains the exclusion phrase "noun" meal "+ verb" make "" in response to a request from the user, for example, "I want to go to a restaurant".
　次に、プリファレンス抽出部１５０は、ステップＳ９２４で得られた除外フレーズについて、視聴行動データベース３００を検索する（ステップＳ９２５）。検索結果として、例えば、以下の情報が得られる。
　　- コンテンツ内容１３１の視聴行動データベース３００内のアドレス Next, the preference extraction unit 150 searches the viewing behavior database 300 for the exclusion phrase obtained in step S924 (step S925). As a search result, for example, the following information can be obtained.
--Address in the viewing behavior database 300 of the content content 131
　そして、プリファレンス抽出部１５０は、ステップＳ９２３で得られた検索結果から、ステップＳ９２５で得られた検索結果に該当するものを除外して、全体の検索結果とする（ステップＳ９２６）。 Then, the preference extraction unit 150 excludes the search results obtained in step S925 from the search results obtained in step S925 to obtain the entire search results (step S926).
　図６は、本技術の実施の形態において除外フレーズの検索結果を除外する例を示す図である。 FIG. 6 is a diagram showing an example of excluding the search result of the exclusion phrase in the embodiment of the present technology.
　同図におけるａは、ステップＳ９２３で得られた検索結果の例である。ここでは、ユーザの要求について、コンテンツＡＡＡＡ、コンテンツＢＢＢＢおよびコンテンツＣＣＣＣなどの検索結果が示されている。 A in the figure is an example of the search result obtained in step S923. Here, search results such as content AAAA, content BBBB, and content CCCC are shown for the user's request.
　同図におけるａは、ステップＳ９２５で得られた検索結果の例である。ここでは、除外フレーズについて、コンテンツＺＺＺＺおよびコンテンツＢＢＢＢなどの検索結果が示されている。 A in the figure is an example of the search result obtained in step S925. Here, search results such as content ZZZZ and content BBBB are shown for the exclusion phrase.
　すなわち、コンテンツＢＢＢＢは、除外フレーズの検索結果に含まれているため、除外される。これにより、同図におけるｃに示すように、コンテンツＢＢＢＢが除外された結果が、全体の検索結果として得られる。 That is, the content BBBB is excluded because it is included in the search result of the exclusion phrase. As a result, as shown in c in the figure, the result excluding the content BBBB is obtained as the overall search result.
　次に、プリファレンス抽出部１５０は、全体の検索結果の中で上位の結果から、コンテンツ内容１３１に紐づいている各メタデータ１３２のスコアを集計する（ステップＳ９２７）。ここで、プリファレンス抽出部１５０は、全体の検索結果の中で上位のコンテンツ内容１３１に対してそれぞれの補正スコアを求める。この補正スコアは、次式により得られる。
　　補正スコア＝スコア×新しさ×視聴回数
ただし、「スコア」は、上述の検索におけるマッチングスコア（尤度）である。また、「新しさ」は、視聴日時が最近のものほど大きい値であり、０から１の値である。 Next, the preference extraction unit 150 aggregates the scores of each metadata 132 associated with the content content 131 from the higher-ranked results in the overall search results (step S927). Here, the preference extraction unit 150 obtains each correction score for the content content 131 that is higher in the overall search results. This correction score is obtained by the following equation.
Corrected score = score x novelty x number of views However, the "score" is the matching score (likelihood) in the above-mentioned search. Further, the "newness" is a value that is larger as the viewing date and time is more recent, and is a value from 0 to 1.
　各コンテンツに紐づいているメタデータのスコアは、紐づいているコンテンツの補正スコアとする。また、同一のメタデータが複数のコンテンツに含まれるときは、合計のスコアとする。例えば、連続ドラマなど繰り返し視聴されるコンテンツに含まれるメタデータは、高いスコアになる。また、連続物ではないが、同じメタデータを含むコンテンツを色々と視聴している場合は、高いスコアになる。 The metadata score associated with each content shall be the correction score of the associated content. When the same metadata is included in a plurality of contents, the total score is used. For example, metadata contained in content that is repeatedly viewed, such as a serial drama, has a high score. Also, if you are watching various contents that are not continuous but contain the same metadata, you will get a high score.
　図７は、本技術の実施の形態においてメタデータのスコアを算出する例を示す図である。 FIG. 7 is a diagram showing an example of calculating a metadata score in an embodiment of the present technology.
　ステップＳ９２６で得られた全体の（除外後の）検索結果について、上述の式から補正スコアを計算する。この例では、コンテンツＡＡＡＡの補正スコア＝７２、コンテンツＣＣＣＣの補正スコア＝１２６、コンテンツＤＤＤＤの補正スコア＝１２０、などが得られる。 For the entire (excluded) search results obtained in step S926, the correction score is calculated from the above formula. In this example, the content AAAA correction score = 72, the content CCCC correction score = 126, the content DDDD correction score = 120, and the like are obtained.
　ここで、メタデータ「食べ物：フランス料理」のスコアを求めるものとする。コンテンツＡＡＡＡおよびコンテンツＤＤＤＤのメタデータには「食べ物：フランス料理」が含まれるため、それぞれの補正スコアが加算される。一方、コンテンツＣＣＣＣのメタデータには「食べ物：フランス料理」が含まれていないため、加算すべき値は０になる。したがって、メタデータ「食べ物：フランス料理」の合計スコアは、７２＋０＋１２０＋…のように計算される。 Here, the score of the metadata "Food: French cuisine" is calculated. Since the metadata of content AAAA and content DDDD includes "food: French cuisine", their respective correction scores are added. On the other hand, since the content CCCC metadata does not include "food: French cuisine", the value to be added is 0. Therefore, the total score of the metadata "Food: French cuisine" is calculated as 72 + 0 + 120 + ...
　次に、プリファレンス抽出部１５０は、上述のように計算されたスコアが上位のメタデータの値をプリファレンスとする（ステップＳ９２８）。これにより、例えば、「フランス料理」、「ミシュラン」、「行列のできるラーメン」などのメタデータの値がプリファレンスとして抽出される。 Next, the preference extraction unit 150 uses the value of the metadata having the higher score calculated as described above as the preference (step S928). As a result, for example, metadata values such as "French cuisine", "Michelin", and "ramen with a queue" are extracted as preferences.
　そして、ユーザ要求処理実行部１６０は、プリファレンス抽出部１５０によって抽出されたプリファレンスに基づいて、ユーザからの要求を処理する（ステップＳ９２９）。例えば、レストランのリストを取得して、プリファレンスへの該当度合いが高い店を有力候補として、ユーザに提示する。上述の例であれば、「フランス料理」、「ミシュラン」、「行列のできるラーメン」の何れかに該当する店がユーザに提示される。 Then, the user request processing execution unit 160 processes the request from the user based on the preference extracted by the preference extraction unit 150 (step S929). For example, a list of restaurants is acquired, and stores with a high degree of appetite are presented to the user as promising candidates. In the above example, the user is presented with a restaurant corresponding to any of "French cuisine", "Michelin", and "Ramen with a line".
　［ニューラルネットワーク］
　図８は、本技術の実施の形態においてユーザ処理要求を受け付ける構成例を示す図である。本技術の実施の形態においては、ユーザからの要求内容を、抽象化表現や除外フレーズに変換するために、学習済みのニューラルネットワークを利用することを想定する。ここでは、ニューラルネットワークの一例として、ＤＮＮ（Deep Neural Network）を想定する。 [neural network]
FIG. 8 is a diagram showing a configuration example for receiving a user processing request in the embodiment of the present technology. In the embodiment of the present technology, it is assumed that a trained neural network is used to convert the content requested by the user into an abstract expression or an exclusion phrase. Here, DNN (Deep Neural Network) is assumed as an example of a neural network.
　ユーザ処理要求受付部１４０は、音声認識器６１０を備える。この音声認識器６１０は、ユーザ要求の音声を音声認識して、その結果としてテキスト形式の出力を生成する。この音声認識器６１０によるテキスト形式のユーザ要求は、ＤＮＮ６２０に供給される。 The user processing request receiving unit 140 includes a voice recognizer 610. The voice recognizer 610 voice recognizes the user-requested voice and, as a result, produces a text output. The text user request by the voice recognizer 610 is supplied to the DNN 620.
　ＤＮＮ６２０は、抽象化表現用のＤＮＮ６２１と除外フレーズ用のＤＮＮ６２２を含む。ＤＮＮ６２１および６２２は、何れも学習済のＤＮＮである。これらＤＮＮ６２１および６２２により、ユーザ要求の抽象化表現および除外フレーズが視聴行動データベース３００に供給され、検索に用いられる。 DNN620 includes DNN621 for abstract expression and DNN622 for exclusion phrase. DNN621 and 622 are both trained DNNs. These DNNs 621 and 622 supply the user-requested abstract expressions and exclusion phrases to the viewing behavior database 300 for use in the search.
　図９は、本技術の実施の形態においてＤＮＮ６２０を学習させる際の構成例を示す図である。 FIG. 9 is a diagram showing a configuration example when learning DNN620 in the embodiment of the present technology.
　ユーザ要求の抽象化のために学習させる場合には、ユーザ要求を入力とし、抽象化表現を出力とする組のデータを蓄積して、教師データ６５０をフィードバックしてＤＮＮ６２０を学習させる。一方、除外フレーズ抽出のために学習させる場合には、ユーザ要求を入力とし、除外フレーズを出力とする組のデータを蓄積して、教師データ６５０をフィードバックしてＤＮＮ６２０を学習させる。なお、学習の初期段階においては、教師データ６５０を手作業により構築することも考えられる。 When learning for abstraction of user request, a set of data in which user request is input and abstract expression is output is accumulated, and teacher data 650 is fed back to train DNN620. On the other hand, when learning for exclusion phrase extraction, a set of data in which a user request is input and an exclusion phrase is output is accumulated, and teacher data 650 is fed back to train DNN620. In the initial stage of learning, it is conceivable to manually construct the teacher data 650.
　このように、本技術の実施の形態によれば、ユーザから処理要求を受けた後に、その要求の特徴に対応するユーザのコンテンツ視聴履歴を検索し、ユーザのプリファレンスを抽出することにより、プリファレンスに基づいてユーザの要求を処理することができる。 As described above, according to the embodiment of the present technology, after receiving a processing request from the user, the user's content viewing history corresponding to the feature of the request is searched and the user's preference is extracted. User requests can be processed based on the reference.
　すなわち、プリファレンスを事後抽出することにより、ユーザからの要求に対応する際に、ユーザの過去のコンテンツ視聴行動に基づいて、目的に合った観点でプリファレンスを抽出することができる。人間の興味は時々刻々と変わり得ることを考慮すると、プリファレンスの事後抽出は合理的である。また、ユーザの過去の視聴行動に遡ったプリファレンス抽出を、ユーザを待たすことなく高速に行うことができる。これにより、ユーザからの要求を、高い精度でユーザの興味の対象に合うように処理できるようになり、ユーザの満足度を向上させることができる。 That is, by extracting the preference after the fact, it is possible to extract the preference from the viewpoint suitable for the purpose based on the user's past content viewing behavior when responding to the request from the user. Post-extraction of preferences is rational, considering that human interests can change from moment to moment. In addition, preference extraction that goes back to the user's past viewing behavior can be performed at high speed without waiting for the user. As a result, the request from the user can be processed with high accuracy so as to meet the interest of the user, and the satisfaction level of the user can be improved.
　＜２．変形例＞
　［購買履歴］
　上述の実施の形態では、コンテンツ視聴履歴に基づいてプリファレンスを事後抽出する例について説明したが、他の行動履歴を利用することも可能である。ここでは、購買履歴を利用する例について説明する。 <2. Modification example>
[purchase history]
In the above-described embodiment, an example of post-extracting the preference based on the content viewing history has been described, but other behavior history can also be used. Here, an example of using the purchase history will be described.
　ユーザが本を購入した際に、その本の内容（あらすじ）をデータベースに登録し、メタデータと紐づける。例えば、「モーツアルトとベートーベンの時代」という本であったとする。その後、そのユーザから、「ＢＧＭを流して」という要求を受けた際に、「曲」「かける」などを検索語にして、データベースを検索する。その検索の結果「モーツアルト」や「ベートーベン」が検索される。これにより、モーツアルトやベートーベンの曲をユーザに提供する。すなわち、音楽を聴くこととは異質な行動履歴から、音楽をかけるためのプリファレンスを抽出することができる。 When a user purchases a book, the content (synopsis) of the book is registered in the database and linked with the metadata. For example, suppose the book was "The Age of Mozart and Beethoven." After that, when the user requests "play BGM", the database is searched using "song", "kake", etc. as search terms. As a result of the search, "Mozart" and "Beethoven" are searched. As a result, Mozart and Beethoven songs are provided to the user. That is, it is possible to extract a preference for playing music from an action history that is different from listening to music.
　［位置情報履歴］
　また、他の行動履歴の例として、位置情報の履歴を利用することも考えられる。例えば、ユーザが携帯する端末のＧＰＳ情報から観光地や商業地などの位置情報を特定して、その特定された位置情報に基づいてユーザの行動内容を文書化してデータベースに登録し、プリファレンスを事後抽出するようにしてもよい。 [Location history]
Further, as an example of another action history, it is conceivable to use the history of location information. For example, location information such as sightseeing spots and commercial areas is specified from the GPS information of the terminal carried by the user, and the user's behavior is documented based on the specified location information and registered in the database, and the preference is set. It may be extracted after the fact.
　［ＤＮＮ］
　上述の実施の形態では、ユーザからの要求内容を、抽象化表現や除外フレーズに変換するためにＤＮＮを利用することを想定したが、これらの変換処理は、ＤＮＮ以外の他の技術により実現してもよい。 [DNN]
In the above-described embodiment, it is assumed that DNN is used to convert the content requested by the user into an abstract expression or an exclusion phrase, but these conversion processes are realized by a technique other than DNN. You may.
　［除外フレーズ］
　上述の実施の形態では、除外フレーズを利用することにより、検索精度を向上させる例について説明したが、検索技術の向上により、ユーザ要求を検索する精度が上がれば、除外フレーズを利用しなくてもよい。すなわち、検索結果のノイズが少なくなれば、除外フレーズは必ずしも利用しなくてもよい。 [Exclusion phrase]
In the above-described embodiment, an example of improving the search accuracy by using the exclusion phrase has been described. However, if the accuracy of searching the user request is improved by improving the search technique, the exclusion phrase may not be used. Good. That is, if the noise of the search result is reduced, the exclusion phrase does not necessarily have to be used.
　［抽象化表現］
　上述の実施の形態では、ユーザの要求を抽象化する例について説明したが、検索技術の向上により、ユーザ要求を検索する精度が上がれば、抽象化表現を利用しなくてもよい。すなわち、ユーザからの要求を直接検索エンジンに入力するようにしてもよい。 [Abstract expression]
In the above-described embodiment, an example of abstracting the user's request has been described, but if the accuracy of searching the user's request is improved by improving the search technique, the abstract expression may not be used. That is, the request from the user may be directly input to the search engine.
　なお、上述の実施の形態は本技術を具現化するための一例を示したものであり、実施の形態における事項と、特許請求の範囲における発明特定事項とはそれぞれ対応関係を有する。同様に、特許請求の範囲における発明特定事項と、これと同一名称を付した本技術の実施の形態における事項とはそれぞれ対応関係を有する。ただし、本技術は実施の形態に限定されるものではなく、その要旨を逸脱しない範囲において実施の形態に種々の変形を施すことにより具現化することができる。 Note that the above-described embodiment shows an example for embodying the present technology, and the matters in the embodiment and the matters specifying the invention in the claims have a corresponding relationship with each other. Similarly, the matters specifying the invention within the scope of claims and the matters in the embodiment of the present technology having the same name have a corresponding relationship with each other. However, the present technology is not limited to the embodiment, and can be embodied by applying various modifications to the embodiment without departing from the gist thereof.
　また、上述の実施の形態において説明した処理手順は、これら一連の手順を有する方法として捉えてもよく、また、これら一連の手順をコンピュータに実行させるためのプログラム乃至そのプログラムを記憶する記録媒体として捉えてもよい。この記録媒体として、例えば、ＣＤ（Compact Disc）、ＭＤ（MiniDisc）、ＤＶＤ（Digital Versatile Disc）、メモリカード、ブルーレイディスク（Blu-ray（登録商標）Disc）等を用いることができる。 Further, the processing procedure described in the above-described embodiment may be regarded as a method having these series of procedures, and as a program for causing a computer to execute these series of procedures or as a recording medium for storing the program. You may catch it. As this recording medium, for example, a CD (Compact Disc), MD (MiniDisc), DVD (Digital Versatile Disc), memory card, Blu-ray Disc (Blu-ray (registered trademark) Disc) and the like can be used.
　なお、本明細書に記載された効果はあくまで例示であって、限定されるものではなく、また、他の効果があってもよい。 It should be noted that the effects described in the present specification are merely examples and are not limited, and other effects may be obtained.
　なお、本技術は以下のような構成もとることができる。
（１）ユーザの要求を取得した後に、前記要求の特徴に対応する前記ユーザの行動履歴から前記ユーザの興味の対象を抽出し、その抽出した前記興味の対象に基づいて前記要求を処理する制御部
を具備する情報処理装置。
（２）前記制御部は、前記要求を抽象化して前記要求の特徴を生成する
前記（１）に記載の情報処理装置。
（３）前記制御部は、ニューラルネットワークを用いて前記要求を抽象化する
前記（２）に記載の情報処理装置。
（４）前記制御部は、前記要求の特徴を生成する際に除外すべき除外フレーズを取得して、前記ユーザの行動履歴から除外する
前記（１）から（３）のいずれかに記載の情報処理装置。
（５）前記制御部は、ニューラルネットワークを用いて前記除外フレーズを取得する
前記（４）に記載の情報処理装置。
（６）前記制御部は、前記ユーザの行動履歴に付随するメタデータのスコアに基づいて前記ユーザの興味の対象を抽出する
前記（１）から（５）のいずれかに記載の情報処理装置。
（７）前記制御部は、前記要求の特徴とのマッチング度合い、前記行動履歴における回数および新しさに基づいて前記メタデータのスコアを生成する
前記（６）に記載の情報処理装置。
（８）前記制御部は、前記スコアが上位となる前記メタデータの値を前記ユーザの興味の対象として抽出する
前記（６）または（７）に記載の情報処理装置。
（９）前記制御部は、前記ユーザの行動内容を前記ユーザの行動履歴として所定のデータベースに登録し、前記ユーザの要求を取得した後に前記データベースから前記ユーザの行動履歴を検索する
前記（１）から（８）のいずれかに記載の情報処理装置。
（１０）前記ユーザの行動履歴は、前記ユーザによるコンテンツの視聴履歴である
前記（１）から（９）のいずれかに記載の情報処理装置。
（１１）制御部が、ユーザの要求を取得する手順と、
　前記制御部が、前記要求の特徴に対応する前記ユーザの行動履歴から前記ユーザの興味の対象を抽出する手順と、
　前記制御部が、前記抽出した前記興味の対象に基づいて前記要求を処理する手順と
を具備する情報処理方法。
（１２）ユーザの要求を取得する手順と、
　前記要求の特徴に対応する前記ユーザの行動履歴から前記ユーザの興味の対象を抽出する手順と、
　前記抽出した前記興味の対象に基づいて前記要求を処理する手順と
をコンピュータに実行させるプログラム。 The present technology can have the following configurations.
(1) Control that after acquiring a user's request, an object of interest of the user is extracted from the behavior history of the user corresponding to the feature of the request, and the request is processed based on the extracted object of interest. An information processing device including a unit.
(2) The information processing device according to (1), wherein the control unit abstracts the requirements and generates features of the requirements.
(3) The information processing device according to (2) above, wherein the control unit abstracts the requirements by using a neural network.
(4) The information according to any one of (1) to (3) above, in which the control unit acquires an exclusion phrase to be excluded when generating the feature of the request and excludes it from the behavior history of the user. Processing equipment.
(5) The information processing device according to (4) above, wherein the control unit acquires the exclusion phrase using a neural network.
(6) The information processing device according to any one of (1) to (5) above, wherein the control unit extracts an object of interest of the user based on a score of metadata associated with the behavior history of the user.
(7) The information processing device according to (6), wherein the control unit generates a score of the metadata based on the degree of matching with the feature of the request, the number of times in the action history, and the novelty.
(8) The information processing device according to (6) or (7), wherein the control unit extracts the value of the metadata having the higher score as the object of interest of the user.
(9) The control unit registers the action content of the user as the action history of the user in a predetermined database, acquires the request of the user, and then searches the action history of the user from the database (1). The information processing apparatus according to any one of (8) to (8).
(10) The information processing device according to any one of (1) to (9) above, wherein the action history of the user is a viewing history of the content by the user.
(11) The procedure for the control unit to acquire the user's request and
A procedure in which the control unit extracts an object of interest of the user from the behavior history of the user corresponding to the feature of the request.
An information processing method in which the control unit includes a procedure for processing the request based on the extracted object of interest.
(12) Procedure for acquiring user's request and
A procedure for extracting an object of interest of the user from the behavior history of the user corresponding to the characteristics of the request, and
A program that causes a computer to perform a procedure for processing the request based on the extracted object of interest.
　１０１　ユーザ視聴要求受付部
　１０２　コンテンツ視聴時処理部
　１０４　ユーザ要求時処理部
　１２０　視聴行動登録部
　１２１、１３１　コンテンツ内容
　１２２、１３２　メタデータ
　１４０　ユーザ処理要求受付部
　１５０　プリファレンス抽出部
　１６０　ユーザ要求処理実行部
　２００　視聴コンテンツ
　３００　視聴行動データベース
　６１０　音声認識器
　６５０　教師データ 101 User viewing request reception unit 102 Content viewing processing unit 104 User request processing unit 120 Viewing behavior registration unit 121, 131 Content content 122, 132 Metadata 140 User processing request reception unit 150 Preferences extraction unit 160 User request processing execution unit 200 Viewing content 300 Viewing behavior database 610 Speech recognizer 650 Teacher data

Claims (12)

  1. 　ユーザの要求を取得した後に、前記要求の特徴に対応する前記ユーザの行動履歴から前記ユーザの興味の対象を抽出し、その抽出した前記興味の対象に基づいて前記要求を処理する制御部
    を具備する情報処理装置。     After acquiring the user's request, the control unit includes a control unit that extracts the object of interest of the user from the behavior history of the user corresponding to the feature of the request and processes the request based on the extracted object of interest. Information processing device.
  2. 　前記制御部は、前記要求を抽象化して前記要求の特徴を生成する
    請求項１記載の情報処理装置。     The information processing device according to claim 1, wherein the control unit abstracts the requirements and generates features of the requirements.
  3. 　前記制御部は、ニューラルネットワークを用いて前記要求を抽象化する
    請求項２記載の情報処理装置。     The information processing device according to claim 2, wherein the control unit abstracts the requirements by using a neural network.
  4. 　前記制御部は、前記要求の特徴を生成する際に除外すべき除外フレーズを取得して、前記ユーザの行動履歴から除外する
    請求項１記載の情報処理装置。     The information processing device according to claim 1, wherein the control unit acquires an exclusion phrase to be excluded when generating the feature of the request and excludes it from the behavior history of the user.
  5. 　前記制御部は、ニューラルネットワークを用いて前記除外フレーズを取得する
    請求項４記載の情報処理装置。     The information processing device according to claim 4, wherein the control unit acquires the exclusion phrase using a neural network.
  6. 　前記制御部は、前記ユーザの行動履歴に付随するメタデータのスコアに基づいて前記ユーザの興味の対象を抽出する
    請求項１記載の情報処理装置。     The information processing device according to claim 1, wherein the control unit extracts an object of interest of the user based on a score of metadata associated with the behavior history of the user.
  7. 　前記制御部は、前記要求の特徴とのマッチング度合い、前記行動履歴における回数および新しさに基づいて前記メタデータのスコアを生成する
    請求項６記載の情報処理装置。     The information processing device according to claim 6, wherein the control unit generates a score of the metadata based on the degree of matching with the feature of the request, the number of times in the action history, and the novelty.
  8. 　前記制御部は、前記スコアが上位となる前記メタデータの値を前記ユーザの興味の対象として抽出する
    請求項６記載の情報処理装置。     The information processing device according to claim 6, wherein the control unit extracts the value of the metadata having the higher score as the object of interest of the user.
  9. 　前記制御部は、前記ユーザの行動内容を前記ユーザの行動履歴として所定のデータベースに登録し、前記ユーザの要求を取得した後に前記データベースから前記ユーザの行動履歴を検索する
    請求項１記載の情報処理装置。     The information processing according to claim 1, wherein the control unit registers the action content of the user as the action history of the user in a predetermined database, acquires the request of the user, and then searches the action history of the user from the database. apparatus.
  10. 　前記ユーザの行動履歴は、前記ユーザによるコンテンツの視聴履歴である
    請求項１記載の情報処理装置。     The information processing device according to claim 1, wherein the behavior history of the user is a viewing history of the content by the user.
  11. 　制御部が、ユーザの要求を取得する手順と、
    　前記制御部が、前記要求の特徴に対応する前記ユーザの行動履歴から前記ユーザの興味の対象を抽出する手順と、
    　前記制御部が、前記抽出した前記興味の対象に基づいて前記要求を処理する手順と
    を具備する情報処理方法。     The procedure for the control unit to acquire the user's request,
    A procedure in which the control unit extracts an object of interest of the user from the behavior history of the user corresponding to the feature of the request.
    An information processing method in which the control unit includes a procedure for processing the request based on the extracted object of interest.
  12. 　ユーザの要求を取得する手順と、
    　前記要求の特徴に対応する前記ユーザの行動履歴から前記ユーザの興味の対象を抽出する手順と、
    　前記抽出した前記興味の対象に基づいて前記要求を処理する手順と
    をコンピュータに実行させるプログラム。     The procedure to get the user's request and
    A procedure for extracting an object of interest of the user from the behavior history of the user corresponding to the characteristics of the request, and
    A program that causes a computer to perform a procedure for processing the request based on the extracted object of interest.
PCT/JP2020/011608 2019-05-27 2020-03-17 Information processing device, information processing method, and program WO2020240996A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2019098210 2019-05-27
JP2019-098210 2019-05-27

Publications (1)

Publication Number Publication Date
WO2020240996A1 true WO2020240996A1 (en) 2020-12-03

Family

ID=73552876

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2020/011608 WO2020240996A1 (en) 2019-05-27 2020-03-17 Information processing device, information processing method, and program

Country Status (1)

Country Link
WO (1) WO2020240996A1 (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2015515674A (en) * 2012-03-15 2015-05-28 セプト システムズ ゲゼルシャフト ミット ベシュレンクテル ハフツングＣＥＰＴ Ｓｙｓｔｅｍｓ ＧｍｂＨ Method, apparatus and product for semantic processing of text
US20150347597A1 (en) * 2014-05-27 2015-12-03 Samsung Electronics Co., Ltd. Apparatus and method for providing information
JP2016184409A (en) * 2010-12-30 2016-10-20 プライマル フュージョン インコーポレイテッド System and method for using knowledge representation to provide information based on environmental input

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2016184409A (en) * 2010-12-30 2016-10-20 プライマル フュージョン インコーポレイテッド System and method for using knowledge representation to provide information based on environmental input
JP2015515674A (en) * 2012-03-15 2015-05-28 セプト システムズ ゲゼルシャフト ミット ベシュレンクテル ハフツングＣＥＰＴ Ｓｙｓｔｅｍｓ ＧｍｂＨ Method, apparatus and product for semantic processing of text
US20150347597A1 (en) * 2014-05-27 2015-12-03 Samsung Electronics Co., Ltd. Apparatus and method for providing information

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
SASAKI, KENTA: "Personal Keyword Extraction from Web Browsing History Exploiting URL Expression and Browsing Frequency", DICOM02010 MULTIMEDIA, DISTRIBUTED, COOPERATIVE, AND MOBILE SYMPOSIUM, vol. 2010, no. 1, 7 July 2010 (2010-07-07), pages 398 - 405 *

Similar Documents

Publication Publication Date Title
US11960526B2 (en) Query response using media consumption history
US9824150B2 (en) Systems and methods for providing information discovery and retrieval
CN111104496B (en) Retrieving context from previous sessions
US8352272B2 (en) Systems and methods for text to speech synthesis
CN109165302B (en) Multimedia file recommendation method and device
US8583418B2 (en) Systems and methods of detecting language and natural language strings for text to speech synthesis
US8712776B2 (en) Systems and methods for selective text to speech synthesis
US8396714B2 (en) Systems and methods for concatenation of words in text to speech synthesis
US20110087549A1 (en) Method and apparatus for providing search capability and targeted advertising for audio, image, and video content over the internet
US20100082328A1 (en) Systems and methods for speech preprocessing in text to speech synthesis
US20090326947A1 (en) System and method for spoken topic or criterion recognition in digital media and contextual advertising
US20210034663A1 (en) Systems and methods for managing voice queries using pronunciation information
JP2011528879A (en) Apparatus and method for providing a television sequence
US20130110513A1 (en) Platform for Sharing Voice Content
CN107145509B (en) Information searching method and equipment thereof
EP3522043A1 (en) Music sharing method and system
JP7171911B2 (en) Generate interactive audio tracks from visual content
KR102340963B1 (en) Method and Apparatus for Producing Video Based on Artificial Intelligence
US20140114656A1 (en) Electronic device capable of generating tag file for media file based on speaker recognition
CN114143479B (en) Video abstract generation method, device, equipment and storage medium
US20210034662A1 (en) Systems and methods for managing voice queries using pronunciation information
JP2019008440A (en) Text extraction apparatus, comment posting apparatus, comment posting support apparatus, reproduction terminal, and context vector calculation apparatus
CN108710653B (en) On-demand method, device and system for reading book
US11410706B2 (en) Content pushing method for display device, pushing device and display device
CN111859008B (en) Music recommending method and terminal

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20815388

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20815388

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: JP