Bidding based on buyer defined functions

Through concept-based bidding methods and advertising processor interpretation technology, the problem of keyword dependence in traditional advertising bidding is solved, more relevant and efficient advertising delivery is achieved, and bid amount and advertising content are dynamically adjusted.

CN114820063BActive Publication Date: 2024-11-26SOUNDHOUND INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210468447.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2016-10-14
Filing Date
2017-09-18
Publication Date
2024-11-26
Estimated Expiration
2037-09-18

AI Technical Summary

Technical Problem

Traditional advertising bids rely on keywords, resulting in long-tail words and keyword query winnings of irrelevant topics. Advertising buyers are unable to effectively consider conceptual relevance and semantic information, resulting in the advertising content being unrelated or not optimized.

Method used

Using concept-based bidding methods, the advertisement buyer specifies the bid amount through functions, uses semantic information and environmental information to dynamically adjust the advertising content, the advertisement processor interprets natural language expressions, generates relevant concept sets, and evaluates the bid amount based on the concept diagram.

Benefits of technology

It improves the relevance of advertising and the efficiency of fund use, allowing advertising buyers to dynamically adjust the bid amount and advertising content according to specific needs, and optimize bidding activities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114820063B_ABST
    Figure CN114820063B_ABST
Patent Text Reader

Abstract

The present disclosure relates to bidding based on buyer defined functions. An advertisement processor evaluates bidding functions based on concepts that may be generated from interpretations of natural language expressions. Ad buyers provide functions with corresponding advertisements to the advertisement processor. The bidding function is also based on the value of semantic information referenced in the expression. The bidding function is also based on environmental information. The advertisement buyer can modify the bidding function. The advertisement may be provided in the form of a question and indicated by an identifying sound. If an expression concept is not found in the bidding function, the expression concept set will be expanded based on the strength of the relationship between the concepts in the concept graph.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Divisional Application Instructions

[0002] This application is a divisional application of the invention patent application with the application date of September 18, 2017, application number 201710842752.6, and name “Bidding based on buyer-defined function”. Technical Field

[0003] The field of the invention relates to advertising bidding in systems performing natural language processing. Background Art

[0004] Traditional advertising relies on keywords. Advertisers specify the bid amount for keywords in order to conduct bidding activities. As a result, in the case of a query using a long-tail word that the bidder did not specify, although the long-tail word expresses a concept related to the advertiser, it will not win the bid. In addition, queries using bid keywords that express a theme that is not related to the advertiser's market will result in irrelevant ads winning.

[0005] Another problem with the traditional approach is that ad buyers can only specify a fixed bid amount for each keyword without considering the relevance of the concept and the specific semantic information associated with the user's expression. Although ad providers have done a lot of research and development for ad buyers to maximize the relevance of content and optimize the efficiency of capital use, ad buyers are still at the mercy of ad providers' keyword and bid amount algorithms. Summary of the invention

[0006] Some, but not necessarily all, novel aspects of the present invention are shown in various embodiments. One aspect is the calculation of advertising bid amounts in the use of concepts compared to keywords. One aspect is that the bids provided by advertising buyers are functions rather than specific amounts. One aspect is the use of one or more expression concepts in a function. One aspect is the use of the value of one or more pieces of semantic information referenced by an expression. One aspect is the use of dynamic values, for example, values ​​that can be accessed through a web application program interface (API). One aspect is the change of advertising content based on the value of the semantic information referenced by the expression. One aspect is the use of one or more environmental information in the bidding function. One aspect is the ability of advertising buyers to modify the bidding function. One aspect is to provide advertisements in the form of questions, for example, advertisements that ask for further requests for information about a product or service. One aspect is to provide audio cues indicating that the verbal response portion is an advertisement. One aspect is to expand the set of expression concepts when it is found that there are no concepts referenced by the bidding function in the set of expression concepts. One aspect is to use a concept map for such expansion. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 Components of a system including an embodiment of the present invention are shown;

[0008] Figure 2 Components of an advertisement processor are shown according to some embodiments of the present invention;

[0009] Figure 3 Components of an advertisement processor with natural language processing and library functions are shown according to some embodiments of the present invention;

[0010] Figure 4 Components of a news system are shown according to some embodiments of the present invention;

[0011] Figure 5 Conceptual diagrams are shown according to some embodiments of the present invention;

[0012] Figure 6 A computer representation of a concept map is shown according to some embodiments of the present invention;

[0013] Figure 7 is a flow chart of a method for generating and using an extended concept set according to an embodiment of the present invention;

[0014] Figure 8 An example of a computing device for implementing the systems and methods disclosed herein is shown. DETAILED DESCRIPTION

[0015] In response to a Texan's request to a natural language-enabled virtual assistant about the weather in New York next week, the assistant reports the weather forecast and then asks the Texan if she needs air tickets and hotels. If so, the assistant provides her with a JetWest airline ticket and a room at the Wyndy Accorn hotel. The assistant provides this service because the Texan is 250 miles away from New York and has plenty of time to fly to New York before next week, and the flight reservation is available at this moment. It may not be appropriate for JetWest and Windy Accorn to bid for their ads using traditional keywords. It is illogical to ask New Yorkers if they need air tickets and hotels in New York. Appropriate natural language concept bids for airlines and hotels should be for expressions about "distant situations". Some expressions can directly indicate concepts, for example, by explicitly mentioning Dallas and New York in a single expression. Some expressions can indicate concepts by immediately referring to New York after an expression about Dallas. Some expressions can implicitly indicate concepts based on the relationship between the expression and information about the person or the environment in which the person is located, for example, an expression from a person known to live in Texas or the fact that the expresser issued the expression in Texas.

[0016] concept

[0017] People can think of concepts as the layout of the connections of neurons in the brain and the discharge patterns of neurons, which can be propagated to form corresponding layouts and patterns in another brain. Words or multi-word (two-word, three-word, n-word) combinations represent concepts. Synonyms are words that represent the same concept, for example, sofa (sofa) and couch (couch). Words in different languages ​​are synonyms, for example, dog (dog) in English and chien (chien) in French. Some words and n-words represent different concepts in different contexts. For example, in the context related to weather, "hot" represents the concept of high temperature, and in the context related to Sichuan cooking, "hot" means high spiciness. Some words or n-words express different concepts simply because of the stress or rhythm when they are spoken. For example, "beautiful big girl" represents different concepts when the stress falls on the word "beautiful" and when the stress falls on the word "big".

[0018] A way to represent a concept in a computer system is to put the concept into an enumeration array or hash list as an index. Basically, each concept is represented by a unique number or address. In various such implementations of the representation concept, the array or hash list contains data indicating the relationship between the words and n words representing each concept and the concept. The association between the concepts forms a concept graph, in which each concept is a node, and the relationship is an edge. In some embodiments, the concept has multiple synonyms or multiple synonym n words, and some embodiments provide contextual information about each word in the n words representing the concept. In some embodiments, the relationship is represented by an intensity value. Basically, each concept has a certain degree of intensity relationship with other concepts, and the intensity degree is very small between most concept pairs. In some embodiments, for actual implementation, only the relationship with an intensity higher than a specific significant degree is represented. For each concept, some such representations are lists of pointers to other concepts. Some embodiments represent the relationship edge of the concept graph as having directionality, and some embodiments have non-directional relationships. Some embodiments represent non-intensity information about concepts, such as the degree of opposition between concepts (love and hate have a high degree of opposition, but mule and donkey have a low degree of opposition), or whether one concept is a component of another concept.

[0019] There are many ways to create a concept map. One way is to search a large number of blog posts or a large number of tweets to determine the dominance based on the proximity of concepts. Another way is to search a dictionary and use the relationship between words in the definitions of other words to identify concept relationships.

[0020] Natural Language Understanding (NLU) is a field that uses computers to analyze utterances to extract concepts and interpret the meaning and intent of the speaker based on the relationships of the concepts. NLU involves many processes, not just simple comparison of keywords or simple mapping of words to enumerated concepts.

[0021] Examples of various aspects

[0022] According to some embodiments of the present invention, the system provides advertisers with concept bidding in place of or in combination with keyword bidding. For example, an expression asking about the weather in Timbuktu next week includes the concept of weather conditions, approximate time (specifically, next week), approximate location (Timbuktu as a specific location), and long distance. Although an ad purchase based on the keyword "weather" would be relevant to such an expression, few ad buyers would bid on enough world place names including Timbuktu.

[0023] According to one aspect of the present invention, the advertisement bidding is for concepts. For example, a hotel management company may bid for the concept of remoteness. The advertisement processor interprets the expression to obtain a set of concepts. For example, the advertisement processor determines whether the expression includes the concept of remoteness according to the position of the expresser. If it does, the advertisement bidding amount is large, otherwise it is not.

[0024] Ad processors implementing some embodiments of the present invention allow ad buyers to specify bid amounts as functions. Bidding functions depend on one or more concepts, and may also depend on semantic information and context information. Some systems support many ads and therefore cannot efficiently check each ad bidding function for each received expression. Some such systems index ads and their associated bidding functions by the concepts used in the function. Thus, for each concept identified in an expression, the system need only directly evaluate the associated bidding function.

[0025] The ad processor evaluates the function based on the concepts in the interpreter's expression, as well as any semantic information or context information referenced in the evaluation function, and then calculates the corresponding bid amount and selects one or more advertisements with the highest bid to be provided to the consumer. For example, a provider of calendar application services may bid on general time concepts. An airline may define a bidding function as follows: an expression with the concept of "next week" corresponds to 2 cents, an expression with the concept of "long distance" corresponds to 20 cents, and an expression containing both concepts corresponds to 1 dollar.

[0026] Advertisement processors implementing some embodiments of the present invention allow ad buyers to specify bidding functions that use values ​​expressed as semantic information. For example, an airline might create a bidding function where the bid amount is equal to 20 cents × log (distance from some place) + 20 cents × (2 days < length of time < 3 months), where "distance from some place" is the distance between two places discernible from the expression, and "length of time" is the time period mentioned in the expression or the time between the current date and the date mentioned in the expression. For purposes of this disclosure, semantic information is information that conveys meaning within an expression. For example, in the sentence "Mary eats noodles," the semantic information is the fact that Mary, eating, and noodles are involved. The sentence has syntax that indicates who is doing something, what is being done, and what is being done to it. For purposes of this disclosure, a semantic information value is a computer representation of specific semantic information.

[0027] An ad processor implementing some embodiments of the invention allows an ad buyer to specify a bidding function based on semantic information. The ad processor receives this information from an appropriate information source and uses it to evaluate the function. The ad buyer can be the source of the data, but the data may also come from a third party, for example, accessing an external application program interface (API). For example, an umbrella supplier may define a bidding function as follows: an expression for the concept of weather corresponds to 15 cents, and the result of a weather information query is a forecast of rain.

[0028] An advertisement processor implementing some embodiments of the present invention allows advertisement buyers to provide variable advertisement content, where the bid amount corresponds to the content. For example, a clothing company may bid on an expression with a weather concept, where a temperature range above 30°C generates a swimsuit advertisement with a high bid amount, a temperature range below 15°C generates a jacket advertisement with a high bid amount, and a temperature between 15°C and 30°C generates a T-shirt advertisement with a low bid amount.

[0029] Advertisement processors implementing some embodiments of the invention allow ad buyers to specify bidding functions using contextual information and semantic information values ​​in addition to the expression concept. Some such contextual information is the time of day, current events, the location and actions of the expresser, information from device sensors, the expresser's behavioral patterns (e.g., activities repeated daily, weekly, and yearly), and the expresser's demographics (e.g., gender, age, race, home location, likes and dislikes, and previous expression history). For example, a clothing supplier may define a bidding function for an ad promoting a swimwear coupon in response to the concepts of travel and water, where the semantic information values ​​are weather above 30°C, and the contextual factors are females between the ages of 15 and 35 years old between 3 and 10 p.m. local time and no more than 15 miles from one of the supplier's retail locations.

[0030] Ad processors implementing some embodiments of the present invention allow ad buyers to modify their bidding functions from time to time. Ad buyers may modify bidding functions for a variety of reasons, such as suspending the use of ads, or weighing the impact of different algorithms on the effectiveness of bidding activities. By controlling the bid amount function, ad buyers can better optimize bidding activities instead of just adjusting keywords and bid amounts and relying on the algorithm of the ad processor.

[0031] Implementation

[0032] Figure 1 Shown is a system comprising an embodiment of the present invention. An expresser 10 is a person who issues a natural language expression. In various embodiments, the expression is a verbal expression, a typing expression or a gesture expression. Other expressions are also feasible. An interpreter component 20 receives the expression, interprets them and outputs a concept set. In certain embodiments, the interpreter is a natural language (NL) processor that receives the expression and outputs the interpretation as a complex data structure. The representation of some complex data structures includes an array that enumerates concept instances, the type of concept in the expression context, the grouping of concepts, and the indication of the relationship between concepts and groups. A content provider 30 uses the interpretation or concept set as the relevant content of input and output. In various embodiments, a content provider is a virtual assistant or other intelligent agent that can receive inquiries and find answers to questions or produce expected behavior, for example, display written or graphic information, output video or audio, or output voice audio with natural language.

[0033] according to Figure 1 In the system, the advertisement processor 40 provides advertisements in response to the expression. Published together with the content are one or more advertisements published by the advertisement processor. These advertisements can be in any form (text, image, video, audio, etc.). The advertisement processor 40 provides advertisements by selecting one or more advertisements from the advertisement database 50. The advertisements in the database 50 come from the advertisement buyers 60. The selection is made based on a comparison of the evaluation results of the bid amount function also provided by the advertisement buyers 60.

[0034] Figure 2 The functionality of the advertisement processor 41 is shown according to an embodiment of the present invention. The expression concept set, semantic information values, and environmental information represent the input data structure of the system. Not shown are external functions for parsing and interpreting the expression to generate the concept set. Also not shown are external functions for accessing data sources to determine semantic information values, such as news headlines, text messages, or other information requested or mentioned by the expresser. Also not shown are external functions for determining environmental information, such as accessing stored user information; location based on location services using a global satellite positioning system; time of day; and other information not mentioned by the expresser.

[0035] Figure 2 The advertisement processor 41 shown in FIG. 4 stores a number of bidding functions 60 defined by advertisement buyers. Each function 60 receives as input a set of expression concepts, a semantic information value, and environmental information. Each function outputs a bidding amount according to an evaluation of its function. The function is a programming function. Any general programming language, for example, C++, Python, or Javascript or any proprietary programming language can be used to implement the present invention.

[0036] Each bid amount corresponds to an advertisement in the advertisement database 50. The advertisement processor compares all bid amounts and selects the bid amount with the highest bid value using a selection function 70. The selector 70 provides an index to the advertisement database 50 to select advertisements from the database for output to the consumer. In some embodiments, the advertisements and corresponding bids are collected in a priority queue. One or more advertisements with the highest bid(s) are selected and the selected advertisements are presented to the consumer.

[0037] Many types of data are interesting to the expressor. Some information is static in nature, e.g., restaurants, Wikipedia articles, and driving directions. Some information is dynamic, e.g., weather, stock prices, news, and sports. Many methods for dynamic data lookup are known. For example, some data sources provide access to their data through a web application programming interface (API), where the API provides data in JavaScript Object Notation (JSON) format from a Hadoop storage device in response to a client uniform resource locator (cURL) access to a representational state transfer (REST) ​​server.

[0038] Some embodiments of the present invention are voice-enabled. Some such systems use automatic speech recognition (ASR) functionality to transcribe the speaker's speech into text. In some embodiments, the ASR functionality is a standalone module, and in some embodiments, it is tightly coupled with natural language processing. In both cases, the ASR functionality transcribes spoken utterances.

[0039] Figure 3An embodiment of the present invention using natural language (NL) processing to generate an interpretation data structure from a text representation of a spoken expression is shown. In this embodiment, the NL processor 80 is part of and controlled by the advertisement processor 42. In other embodiments with NL processing, the NL processor is an independent module. The NL processor interprets the expression as an input and outputs an interpretation data structure. Some embodiments use neural networks and rely on word embedding. In some embodiments, the NL processor interprets the expression, parses the expression, and uses syntax-oriented semantics to build an interpretation according to a semantic grammar. Some such embodiments build an interpretation of the expression and then extract a concept set from the interpretation. Some NL processors also return an interpretation data structure and are further processed on the behalf of the expresser, for example, by making a dictation transcript or answering a question. Some embodiments, when encountering a response to a question expression, add the concept in the answer to the interpretation or concept set.

[0040] Fully parsing some expressions requires more than one semantic information value. For complex expressions, some information lookups use additional information values. Values ​​can be anything that can be expressed as a number, including numbers belonging to enumeration types, indices into lists or arrays, database record numbers, and pointers to other values. Values ​​can represent any type of information, including information such as weather conditions, location, or time.

[0041] Figure 3 The use of library functions 90 is also shown. In order to simplify programming and make the ad bidding function available to a large number of advertisers, some ad processors provide software development kits (SDKs) to ad buyers. The SDK provides a set of library functions 90 that operate on complex interpretation data structures, specific semantic information, and various environmental information to save the ad buyer from having to understand all the details. For example, an SDK library function provides the distance from the current location. For each specific location mentioned in the expression, the SDK returns the difference between the location of the specific location and the GPS coordinate location of the consumer (if known). When the search exceeds a certain threshold distance, such as 200 miles, the search can be classified as "long distance". Another SDK library function receives a specific time value and location value, executes a web application (API) accessing a weather forecast provider, and returns the expected high temperature. Another SDK library function receives a consumer ID, checks the consumer's Facebook profile, and returns the consumer's age, gender, marital status, and zip code.

[0042] Some embodiments use an editor to compile the bidding function into code in an executable computer assembly language. The code is compact enough that it can run thousands of ads for each expression from a cache or fast local memory without accessing mass storage. Some embodiments store the bidding function in an interpreted language that allows ads to be added and removed dynamically without causing delays when recompiling large groups of ads. Some embodiments store the bidding functions sequentially based on a simple order, such as when they are sorted by ad buyers or alphabetically by name. Some embodiments sort the bidding functions based on possible run times so that faster bidding functions run first. Some embodiments index the bidding functions by one or more specific concepts, library functions, types of semantic information, and types of environmental information referenced by the bidding function. For each expression, the system only evaluates the bidding functions, library functions, types of semantic information, or types of environmental information that mention the specific concept of the expression.

[0043] According to some embodiments of the present invention, different devices implement Figure 1 The expresser 10, the function 20 for interpreting the expression, the advertisement processor 40, the advertisement database manager 50, the advertisement buyer 60, and the content provider 30 may be separate entities, or any specific entity that may include multiple components.

[0044] Advertising

[0045] Once the advertisement is selected and dynamically configured (if necessary), it must be delivered to the consumer. Various natural language processing systems may include embodiments of the present invention. Some such systems are natural language search engines, text-based automated agents, language-based automated agents, spoken conversational virtual assistants, robot assistants, voice-enabled devices, vehicles such as cars and airplanes, retail kiosks, media entertainment systems, and voice-enabled applications for mobile devices. Some systems including embodiments of the present invention accept spoken expressions and provide natural language responses. Natural language responses include relevant content and voice natural language advertisements. Some systems also provide visual or video advertisements or provide visual or video advertisements in place of voice responses.

[0046] Some embodiments of the present invention provide advertisements by asking questions or statements to the expresser to prompt subsequent expressions. For example, a yogurt manufacturer may provide an advertisement such as "Our yogurt is particularly nutritious, would you like to learn more about the benefits of probiotics?" Such an advertisement processor system will recognize subsequent expressions such as "Yes" or "Tell me more" as a direct request for a second advertisement segment from the yogurt supplier.

[0047] Some embodiments of the present invention provide audio cues that indicate an advertisement. The cue can be a simple beep or a unique sound. The audio cue can enable the presenter to distinguish specific query response content from unsolicited advertising content. Doing so can improve the usefulness of certain auxiliary systems.

[0048] application

[0049] The present invention is useful for various types of advertising. The present invention can be used for online and mobile advertising of major brand products and services. It can also be used for political advertising. Some embodiments store a personal history of each utterer. Some embodiments store the history as utterance text. Some embodiments store the history as an array of keywords and their usage counts, and then process the keywords as concepts for bid amount functions. Some such embodiments periodically decrement the counts so that old utterances gradually fade out. Some embodiments represent the history as concepts rather than keywords.

[0050] Some embodiments allow political ad buyers to employ finer granularity to replace non-traditional demographic-based targeting in order to tailor their ads to cover specific individuals' buying interests. Such ad targeting is not limited to individuals, but allows ads to be targeted to their specific interests. In addition, analysis of expressions posted after the ad is placed helps determine persuasiveness and predict voting preferences. With the proliferation of devices that collect expressions, this function-based ad buying provides greater coverage than traditional methods.

[0051] Another application of the present invention for advertising bidding and placement is in the field of print and mail advertising, such as newspapers, magazines, and coupon flyers. According to an embodiment of the present invention, a coupon flyer publisher, which acts as an advertising processor, allows buyers to create advertising functions. The publisher uses a geographic concept map to identify concepts expressed in communities and individual homes. The publisher automatically places ads within the flyer based on the ad buyer's functions, sets ad sizes, sorts ads, and customizes them for each address.

[0052] Figure 4 Another embodiment of the present invention for newspaper and magazine print advertising is shown. Publisher 31 receives stories from story writers 14 and provides content to readers 15. Publisher 31 also provides content to advertisement processor 43. Advertisement processor 43 uses a bidding function from advertisement buyer 61 to select advertisements from advertisement database 50 and sends the selected advertisements to readers 15.

[0053] According to the present invention, as a large number of mainstream publications adopt variable data printing, publishers 31 can become advertising processors 43. They can provide advertising buyers 61 with the option of defining functions based on demographic and other specific profile information of the collection of reading subscribers 15. The bidding amount function of advertising buyers 16 can also take into account the content of the publication, such as weather forecasts, stock prices, and concepts expressed in the story. In such an embodiment, the story author 14 is the expresser, and the publication reader 15 is the recipient of the advertisement. This is different from some other embodiments in that the expresser of the expression and the consumer of the advertisement are different people.

[0054] Concept map

[0055] If the ad processor receives the expression: "I'm busy right now (tied up), a purchase ad based on the keywords simple rope or scissors will have a high bid amount. This will annoy the expression buyer with irrelevant ads and waste the ad buyer's money. According to some embodiments of the present invention, the concept-based system for determining ad bid amounts will recognize that the expression does not include the concept of tied up, but rather the concept of busy. Ad bids for time management software or personal assistant services will have higher bid amounts.

[0056] If an ad processor receives an expression containing the acronym: "Find me a solution for ERP?", various ads with inappropriate bid amounts may be generated due to the ambiguity of the acronym ERP. According to some embodiments of the present invention, an ad processor providing a bid function that takes personal profile information as input will recognize that for a business executive expression, ERP refers to Enterprise Resource Planning, but for an event organizer, ERP refers to Emergency Response Plan.

[0057] If the advertisement processor receives the expression: "Where was garlic first cultivated?" and no manufacturer purchases the advertisement, all advertisement bidding functions are likely to offer low bid amounts. Due to the lack of a strong bidding signal, the selection of advertisements is equivalent to being based on noise, and the selected advertisements are often irrelevant. According to some embodiments of the present invention, a concept-based advertisement bidding system using a bidding function provides improved results by weighting the results of the function based on the strength of the relationship between the concept of the expression and different concepts in the function. Even if no bid amount is assigned to the concept of garlic, the concept of garlic has a strong relationship with the concept of elements in Gothic novels. Another bidding function for advertisement purchases by a bookstore that sells Gothic novels bids on the concept of elements in Gothic novels and will offer a relatively high bid amount even if its function does not include the concept of garlic. For this expression, the concept of garlic is included, which has a strong relationship with the concept of food, which in turn has a strong relationship with the concept of gardening, and the expression includes the concept of cultivation, which has a strong relationship with home, which in turn has a strong relationship with garden, so even if the advertisement bidding function does not include the concept of garlic, the garden supplier will offer a relatively high bid amount.

[0058] Some embodiments of the present invention provide an advertising bidding function that determines a bid amount based on the relationships between concepts and the strengths between those relationships determined using a concept graph. Some embodiments of the present invention represent a concept graph as a two-dimensional array of strength values ​​between each concept and all other concepts. Many methods for representing concept graphs are known and are applicable to various embodiments of the present invention. Some concept graphs are directed, some are non-directed. Some include information other than strength between nodes, such as the type of relationship.

[0059] Figure 5 shows a small concept graph, where concepts are nodes with edges showing the strength of the relationship. Concepts are nouns, verbs, and modifiers, which can be named n, v, and m respectively. A particular concept in the graph can be represented by multiple words. It is not necessary for all words to be synonyms, only that they invoke the same concept in some context. Figure 5 It is appropriate to list multiple words for a concept node.

[0060] Figure 6 A specific representation of a concept map in a computer file is shown. First is a concept list segment. Each concept has a letter and a number, the letters representing a noun, a verb, and a modifier, respectively, and each number is unique within the same concept type. A colon and a space following each concept identifier separate a list of words, expressing the concept within quotation marks. This embodiment of the present invention ignores each type of preposition, article, and other words that are not nouns, verbs, modifiers, or phrases.

[0061] In the document, the second part is a list of concept pairs indicated by joint concept identifiers. A colon and a number after each pair of concepts indicate the strength of the relationship between the two concept pairs.

[0062] Consider the concepts "car", "parts", and "repair" in a database of auto parts supplier demand advertisements. These concepts are n0, n5, and v1 in the concept graph.

[0063] Consider the expression: "My radiator is broken", which is interpreted as having concepts n4 and m0. It has no keyword matches with the advertisement. Some embodiments of the present invention expand the set of concepts to include adjacent concepts. Some embodiments expand starting with the least frequent expression concept, which for this expression is n4 radiator. The set is expanded to include n0, n1, n5, and v1 in addition to n4 and m0. Some embodiments expand starting with the original expression concept with the most relationships. Some expand starting with the concept with the least relationships. Some expand starting with the concept with the greatest concept pair strength. Various algorithms are suitable. The expanded set of expression concepts includes n0 and v1, both of which overlap with the advertisement concept, so that an advertisement is selected and presented to the user.

[0064] Some embodiments search for bid functions that reference any concept in the set across all or a set of bid functions. Some such embodiments iteratively perform expansion until at least one matching concept is found between the concept referenced in the bid function and the expanded set of expressed concepts. If multiple matching concepts are found, the concept relationship strength is used to select the best advertisement. In some embodiments, expansion is performed until a certain number of matching concepts are found, and then the relationship strength is calculated to select the best advertisement.

[0065] Consider the same scenario, but involving another advertisement for a car towing company requiring the concepts "car" and "damage". That is, the two concepts are n0 and m0. The user expression interpreted with concepts n4 and m0 matches concept m0 of the towing company's advertisement, so it is presented to the user before the concept set is expanded.

[0066] Consider the third advertisement involving the concepts of "car", "damage" and "repair" required by a car repair shop. That is, the three concepts are n0, v1 and m0. The user expressed the concept m0 that matches both the towing company and the car repair shop. Some embodiments calculate the average relationship strength between the connections. The default value for the average relationship strength between the expressed concept and the concept of the towing company advertisement is 4.5. The default value for the average relationship strength between the expressed concept and the concept of the repair shop advertisement is 6.5. The average value for auto parts and accessories suppliers is also 6.5. This is mainly because of the relationship strength between the concepts of damage and repair. Since the repair shop provides repair services, while the towing company does not, the repair shop advertisement is presented.

[0067] The average strength is better than the sum strength because the average gives advertisers an incentive to choose relevant concepts rather than just the majority concept.

[0068] Considering the demographic overlay based on gender, all repair concept relationships are reduced by one-half, and all radiator concepts that are relevant to females are reduced to one-tenth relative to the male demographic. This results in a default average relationship strength of only 3.675 between the expression concept and the auto repair shop and parts supplier concepts. Therefore, the towing company ads are presented to females, while the repair shop and parts supplier ads are presented to males.

[0069] Consider the second expression: "There is a hole and the antifreeze is leaking", which is interpreted as containing concepts n1, n3, and v0. When compared with towing service, repair shop, and parts supplier, respectively, these concepts have an average relationship strength default value of about 7.5, 7.3, and 7.0. Therefore, in response to the second expression, the towing service advertisement will be presented.

[0070] Consider the overlay of geographic location. If the location is close to a repair shop, an ad for the repair shop will pop up. Consider the overlay of the user's proximity to home. If the user's proximity to home is high, an ad for a car parts supplier will be presented. Consider aspects regarding the time of day. If the time is after hours, the repair shop will decrease, while a 24-hour towing service will not. Similarly, the relationship strength for antifreeze will increase in the winter and decrease in locations with warmer weather.

[0071] Overlays may be generated based on factors related to the grammar of the expression. For example, concepts expressed as subjects, predicates, or objects in an expression may increase the strength of the relationship between them, while concepts expressed in subordinate clauses may weaken the strength of their relationship.

[0072] Some embodiments of the present invention weight the strength of concept pairs based on the parts of each speech. In some embodiments, the relatively important parts of speech are sorted as nouns, verbs, and then modifiers. In some embodiments, the relatively important parts of speech are sorted as verbs, nouns, and then modifiers. Some embodiments of the present invention weight the strength of concept pairs according to where each concept appears in the grammatically meaningful sentence in the expression. In particular, subject-predicate concept pairs and predicate-object concept pairs have greater weights than concept pairs involving clauses and modifying phrases.

[0073] Many methods for creating, training, and evolving concept graphs are known in fields such as machine learning and artificial intelligence. Most known methods can be used to construct concept graphs suitable for the present invention. Many methods involve training concept graphs on appropriate corpus data. According to some embodiments of the present invention, an ad receiver receives specific concept graphs provided by ad buyers. They are motivated to create large and meaningful concept graphs in order to improve the relevance of their ads (thereby improving the efficiency of the use of funds). The ad processor aggregates the ad buyers' concept graphs into a common concept graph. As the amount of ad purchases increases, the ad processor's common concept graph improves accordingly.

[0074] According to some embodiments of the invention, to accommodate evolving language, the ad processor gradually reduces the strength of concept connections over time, increasing them only when ad buyers repeatedly provide and reinforce the strength of the connections. To accommodate short-term fads, the ad processor looks for concept connections that have a large difference in strength between old and new concept graphs. Stronger weights are used, but they are reduced rapidly over time until they reach medium or low strength.

[0075] Some embodiments construct an original concept set by parsing an input expression. Various embodiments represent the concept set as an array of strings, an array of Boolean instances of an enumeration type, a data structure indicating relationships between concept instances, a data structure indicating the strength of concepts within an expression, and various other appropriate representation structures. Some embodiments retain knowledge (e.g., in list form) of the concepts mentioned in each bidding function. Some embodiments index bidding functions by concept for fast access. Some embodiments use concepts in the original concept set in response to the expression to access only the indexed bidding functions for evaluating the advertising bid amount.

[0076] Some embodiments, when it is found that no concept in the bidding function matches a concept in the original concept set, use intelligently expanding the concept set in an iterative manner to include other concepts, and searching the advertisement database or an index thereof to base the bidding function on at least one concept in the concept set. According to a specific embodiment, Figure 7 Such a method is shown. Process 700 begins when the raw set forming step 701 receives the expression analysis. Step 701 generates a raw concept set for subsequent processing. Step 702 receives the raw concept set and searches in the advertisement data index 703 for each concept in the concept set to identify whether there is any advertisement bidding function based on any concept in the concept set. The advertisement database index 703 stores the index of all concepts referenced in each bidding function.

[0077] Some embodiments intentionally search for more than one bid function that includes concepts that match the concept set. This helps ensure that the bid amount calculation is evaluated, and although the bid function does not directly match the original concept set, it may produce a larger bid amount than a bid function that matches the original or expanded concept set.

[0078] If step 702 finds that there are less than a target number of bid functions in the index that match concepts in the concept set, processing continues with step 704 where the set is expanded. Many algorithms are known for expanding a set of nodes based on the attribute values ​​between nodes. Any such algorithm is an embodiment of the present invention, and a specifically designed set expansion algorithm is part of how an ad processor differentiates the ads it provides by targeting a specific application. Some examples that are considered suitable for concept set expansion are the order of strength of potential additional concepts, the distance of potential additional concepts from the original concept set, the connection strength accumulated through multi-graph routes from the concept set and the potential additional concepts, and the connection strength accumulated from each concept set to each potential additional concept through multi-graph routes. Some algorithms add one concept to the concept graph per iteration. Some algorithms add multiple concepts to the concept graph per iteration.

[0079] After the step of creating the expanded concepts, the process returns to step 702. If step 702 finds at least the target number of bid functions that match the concepts in the set in the index, the process proceeds to step 705, where the system performs a process of evaluating each matching bid function to determine the concept with the highest bid value. Next, the process continues to step 706, where the system obtains from the advertisement database 707 an advertisement corresponding to the bid function that is evaluated to produce the maximum bid value. Step 706 completes the process by providing the advertisement to the consumer.

[0080] Figure 8 800 is a block diagram depicting an example of a computing device 800. The computing device 800 can perform, for example, similar tasks and operations discussed herein regarding determining network data to be displayed, generating network diagrams, implementing various algorithms, etc. The computing device 800 can be used to perform various steps as discussed herein. The computing device 800 can be used as a server, a client, or any other computing entity. The computing device 800 can be any of a variety of computing devices, for example, a desktop computer, a notebook computer, a server computer, a handheld computer, a tablet computer, a mobile device, a set-top box, a portable entertainment device, and similar computing devices.

[0081] The computing device 800 includes one or more processors 802, one or more memory devices 804, one or more interfaces 806, one or more mass storage devices 808, and one or more input / output (I / O) devices 810, all coupled to a bus 812. The one or more processors 802 include one or more processors or controllers that execute instructions stored in the one or more memory devices 804 and / or the one or more mass storage devices 808. The one or more processors 802 may also include various types of computer-readable media, such as cache memory.

[0082] The one or more storage devices 804 include various computer-readable media, such as volatile memory (e.g., random access memory (RAM) 814) and / or non-volatile memory (e.g., read-only memory (ROM) 816). The one or more storage devices 804 may also include a rewritable ROM such as flash memory.

[0083] The one or more mass storage devices 808 include various computer-readable media, such as tapes, disks, solid-state memory (e.g., flash memory), etc. Various drives may also be included in the one or more mass storage devices 808 to facilitate reading from and / or writing to various computer-readable media. The one or more mass storage devices 808 include removable media 826 and / or non-removable media.

[0084] The one or more I / O devices 810 include various devices that allow data and / or other information to be input to or retrieved from the computer device 800. Examples of the one or more I / O devices 810 include cursor control devices, keyboards, keypads, microphones, monitors or other display devices, speakers, printers, network interface cards, modems, lenses, charge coupled devices (CCDs) or other image acquisition devices, and the like.

[0085] The one or more interfaces 806 include various interfaces that allow the computing device 800 to interact with other systems, devices, or computing environments. Examples of the one or more interfaces 806 include any number of different user interfaces 818, such as a touch screen, keyboard, mouse, joystick, and remote control, network interfaces 820 such as interfaces to local area networks (LANs), wide area networks (WANs), wireless networks, and the Internet, and peripheral device interfaces 822 such as USB, RS-232, parallel port, and high-definition multimedia interface (HDMI).

[0086] The bus 812 allows the one or more processors 802, the one or more storage devices 804, the one or more interfaces 806, the one or more mass storage devices 808, and the one or more I / O devices 810 to communicate with each other and with other devices or components coupled to the bus 812. The bus 812 represents one or more of several types of bus structures, such as a system bus, a PCI bus, an IEEE 1394 bus, a USB bus, etc.

[0087] It should be understood that such programs and components may be stored in different storage components of the computing device 800 at different times and executed by one or more processors 802. Alternatively, the systems and programs described herein may be implemented in hardware, or a combination of hardware, software, and / or firmware. For example, one or more application specific integrated circuits (ASICs) may be programmed to execute one or more of the systems and programs described herein.

[0088] Although the present disclosure is described by certain preferred embodiments, other embodiments are obvious to those skilled in the art, and considering the benefit of the present disclosure, embodiments including those that do not provide all the benefits and features described above are also within the scope of the present disclosure. It should be understood that other embodiments can be used without departing from the scope of the present disclosure.

[0089] In the above disclosure, the accompanying drawings are described, which form a part of the present disclosure and are shown by describing specific embodiments for implementing the present disclosure. It is understood that other embodiments can be used and structural changes can be made without departing from the scope of the present disclosure. "One embodiment", "embodiment", "example embodiment" and the like mentioned in the specification indicate that the described embodiment may include specific features, structures, or characteristics, but each embodiment does not necessarily include these specific features, structures, or characteristics. In addition, these phrases do not necessarily refer to the same embodiment. In addition, when specific features, structures, or characteristics are described in conjunction with an embodiment, it is considered that such features, structures, or characteristics can be combined with other embodiments within the scope of the knowledge of those skilled in the art, whether or not they are explicitly described.

[0090] The embodiments of the systems, devices and methods disclosed herein may include a special-purpose computer or a general-purpose computer, which includes computer hardware, such as one or more processors and system memory discussed herein. Embodiments within the scope of the present disclosure may also include physical media and other computer-readable media for carrying or storing computer-executable instructions and / or data structures. Such computer-readable media may be any available media accessible by a general-purpose computer system or a special-purpose computer system. The computer-readable medium storing computer-executable instructions is a computer storage medium (device). The computer-readable medium carrying computer-executable instructions is a transmission medium. Therefore, by way of example and not limitation, the embodiments of the present disclosure may include at least two obviously different types of computer-readable media: a computer storage medium (device) and a transmission medium.

[0091] Computer storage media (devices) include RAM, ROM, EEPROM, CD-ROM, solid-state drives ("SSD") (e.g., RAM-based SSDs), flash memory, phase-change memory ("PCM"), other memory types, other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other media that can be used to store desired program code in the form of computer-executable instructions or data structures and that can be accessed by a general purpose or special purpose computer.

[0092] The embodiments of the device, system and method disclosed herein can communicate via a computer network. "Network" is defined as one or more data links that facilitate electronic data to be transmitted between a computer system and / or module and / or other electronic device. When information is transmitted or provided to a computer via a network or other communication connection (hardwired, wireless or a combination of hardwired or wireless), the computer properly regards the connection as a transmission medium. Transmission media may include a network and / or a data link, which can be used to carry a desired program code in the form of a computer executable instruction or data structure, and can be accessed by a general-purpose computer or a special-purpose computer. The above combination should also be included in the scope of computer-readable media.

[0093] Computer executable instructions include, for example, instructions and data that cause a general-purpose computer, a special-purpose computer, or a special-purpose processing device to perform a function or group of functions when executed by a processor. Computer executable instructions may be, for example, binary instructions, intermediate format instructions such as assembly language, or even source code. Although the subject matter has been specifically described in language with respect to structural features and / or method actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the features or actions described above. Instead, the described features and actions are disclosed as example forms of implementing the claims.

[0094] Those skilled in the art will appreciate that the present disclosure can be implemented in a network computing environment with many types of computer system configurations, including built-in vehicle computers, personal computers, desktop computers, laptop computers, message processors, hand-held devices, multiprocessor systems, microprocessor-based electronic products or programmable consumer electronics, network PCs, minicomputers, mainframe computers, mobile phones, personal digital assistants, tablet devices, pagers, routers, switches, various storage devices, etc. The present disclosure can also be implemented in a distributed system environment, where local computer systems and remote computer systems linked by a network (by a hardwired data link, a wireless data link, or a combination of hardwired and wireless data links) all perform tasks. In a distributed system environment, program modules can be located in local storage devices and remote storage devices.

[0095] In addition, the functions described herein may be performed in one or more of hardware, software, firmware, digital components, or analog components, where appropriate. For example, one or more application specific integrated circuits (ASICs) may be programmed to perform one or more of the systems and processes described herein. Certain terms are used throughout the description and claims to refer to specific system components. Those skilled in the art will appreciate that components may be referred to by different names. This document is intended to distinguish between components with different functions, without distinguishing between components with different names.

[0096] It should be noted that the above sensor embodiments may include computer hardware, software, firmware, or any combination thereof to perform at least a portion of their functions. For example, the sensor may include computer code configured to be executed in one or more processors, and may include hardware logic / circuitry controlled by the computer code. These example devices provided herein are for illustrative purposes and are not intended to limit the present invention. The embodiments of the present disclosure may be implemented in other types of devices known to those skilled in the relevant art.

[0097] At least some embodiments of the present disclosure are directed to a computer program product containing such logic stored on any computer usable medium (e.g., in the form of software). Such software, when executed in one or more data processing devices, causes the devices to operate as described herein.

[0098] The following is not intended to be examined as a claim. It is a non-essentially expanded version of the required claim. This version is a potential alternative only if needed to overcome a rejection due to the brevity of the required claim:

[0099] A method for providing an advertisement to a user of a natural language processing system, the method comprising:

[0100] Interpret user queries using natural language processing;

[0101] generating an advertisement in response to the query; and

[0102] providing a spoken response in response to the query, the response including the advertisement,

[0103] The advertisement prompts the user to request goods or services in the form of questions.

Claims

1. A method for providing an advertisement to a user of a natural language processing system, the method comprising: Interpret user queries using natural language processing; generating an advertisement in response to the query; as well as providing a spoken response in response to the query, the response including the advertisement, The advertisement prompts the user to request goods or services in the form of a question. Wherein, the method further comprises: outputting a concept set based on the interpretation of the query; receiving an answer to the question; and Add the concept in the answer to the concept set, The question is related to the concept included in the query.

2. The method according to claim 1, further comprising: Written or graphical information, or a video is provided in response to the query.

3. The method according to claim 2, further comprising: A text advertisement, an image advertisement, or a video advertisement is generated in response to the query.

4. A device for providing advertisements to a user of a natural language processing system, the device comprising: means for interpreting user queries using natural language processing; means for generating an advertisement in response to said query; as well as means for providing a spoken response in response to said query, wherein said response includes said advertisement, The advertisement prompts the user to request goods or services in the form of a question. Wherein, the device also includes: means for outputting a set of concepts based on the interpretation of the query; means for receiving an answer to said question; and means for adding concepts in said answer to said set of concepts, The question is related to the concept included in the query.

5. The device according to claim 4, further comprising: Means for providing written or graphical information, or a video, in response to the query.

6. The device according to claim 5, further comprising: Means for generating a text advertisement, an image advertisement, or a video advertisement in response to the query.