WO2020000867A1 - 一种答案提供方法及设备 - Google Patents

一种答案提供方法及设备 Download PDF

Info

Publication number
WO2020000867A1
WO2020000867A1 PCT/CN2018/115430 CN2018115430W WO2020000867A1 WO 2020000867 A1 WO2020000867 A1 WO 2020000867A1 CN 2018115430 W CN2018115430 W CN 2018115430W WO 2020000867 A1 WO2020000867 A1 WO 2020000867A1
Authority
WO
WIPO (PCT)
Prior art keywords
answer
user
type
question
content
Prior art date
Application number
PCT/CN2018/115430
Other languages
English (en)
French (fr)
Inventor
缪庆亮
胡长建
徐飞玉
戴炜
何海洋
Original Assignee
联想(北京)有限公司
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 联想(北京)有限公司 filed Critical 联想(北京)有限公司
Publication of WO2020000867A1 publication Critical patent/WO2020000867A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue

Definitions

  • the present disclosure relates to the field of electronic devices, and more particularly, the present disclosure relates to an answer providing method and device.
  • a domain expert needs to manually judge the corresponding answer content to set a unique answer in advance. Furthermore, when the user enters the question, the customer service system will push the user with a preset unique answer. However, the content of the pre-set unique answer is fixed and of a single type, which cannot meet the different needs of the user, resulting in a decrease in user satisfaction with the push answer provided by the customer service system. On the other hand, the process of setting a unique answer in advance requires a domain expert to make a manual judgment on the type of answer, which costs a lot of manual labeling costs.
  • an answer providing method which includes: receiving a question input by a user; determining a first answer corresponding to the question input by the user; A first answer is provided to the user, wherein the question also corresponds at least to a second answer, and the second answer is an answer provided to the user in different question input modalities.
  • an answer providing device including: a question receiving unit configured to receive a question input by a user; an answer determining unit configured to determine a first answer corresponding to a question input by a user; and an answer A providing unit is configured to provide the first answer to a user, wherein the question further corresponds to at least a second answer, and the second answer is an answer provided to the user in different question input modalities.
  • a computer-readable storage medium for storing non-transitory computer-readable instructions that can be executed when the non-transitory computer-readable instructions are executed by a computer. Answers provide methods.
  • a computer program including computer-executable instructions that can execute the answer providing method of an embodiment of the present disclosure when the computer-executable instructions are executed.
  • a corresponding answer can be determined based on a question input by a user and different answers can be provided to the user in different modalities, thereby meeting various needs of the user for the answer, Increase user satisfaction with answers.
  • FIG. 1 is a flowchart illustrating an answer providing method according to an embodiment of the present disclosure
  • 2A and 2B are exemplary methods illustrating determining a first answer corresponding to a question input by a user according to an embodiment of the present disclosure
  • FIG. 3 is an exemplary method illustrating the type of answer used to represent an answer based on the content of the answer according to an embodiment of the present disclosure
  • FIG. 4 is an exemplary method illustrating determining an answer type to be used to represent an answer based on the respective priority scores according to an embodiment of the present disclosure
  • 5 is an exemplary method illustrating whether to construct an answer represented by an answer type having the highest priority score according to an embodiment of the present disclosure
  • 6A is a flowchart illustrating an exemplary method of constructing a step cost knowledge base according to an embodiment of the present disclosure
  • 6B is a schematic diagram illustrating a step cost knowledge base constructed according to an embodiment of the present disclosure
  • FIG. 6C is a schematic diagram illustrating a step weighted construction cost of each operation step in a step cost knowledge base according to an embodiment of the present disclosure
  • FIG. 7 is a schematic diagram illustrating matching of a historical question and a standard question of a user according to an embodiment of the present disclosure
  • FIG. 8 is a block diagram illustrating an answer providing apparatus according to an embodiment of the present disclosure.
  • the answer providing method and device are described using the customer service system in the mobile phone field as an example of a question answering system. Those skilled in the art can understand that the answer providing method and device of the present disclosure can also be applied to question answering systems in other fields. For the convenience of explanation and understanding, the following are examples of the content of several standard questions and answers in the field of mobile phones for reference.
  • the bootloader can be unlocked.
  • the bootloader is a piece of code that runs when the device is verified to verify that the software is official. Unless you have experience in this area, we do not recommend unlocking the bootloader.
  • the answers pushed to users are fixed in content and of a single type.
  • the existing mobile customer service system only provides a unique answer with a single type and fixed content, so as to push the unique answer to all users who ask the question.
  • a certain user input modality may reflect that the user only supports listening to voice-type answers to learn certain operation methods (for example, learning to set up a VPN Operation method), but the existing customer service system may only push the answer of the video type; for example, another user input modal may reflect that the user wants a more concise / exhaustive answer to the technical terms (for example, grasp the meaning of the bootloader) , But existing customer service systems push answers to all users with exactly the same level of detail. Therefore, the only answer pushed by the existing customer service system to the user may not be suitable for different needs of different users and / or the same user, and cannot well meet the user needs in different question input modalities.
  • an embodiment of the present disclosure proposes an answer providing method, which determines and provides an answer corresponding to a question input by a user, and can provide a user with different answers in different question input modalities, thereby Push the right answer to the user.
  • an answer providing method according to an embodiment of the present disclosure will be described with reference to FIG. 1. As shown in FIG. 1, the answer providing method of this embodiment may include the following steps:
  • step S101 a question input by a user is received.
  • a user can use his or her electronic device to input questions to the customer service system in various ways.
  • the manner in which the user enters the question may include, but is not limited to: typing input, voice input, selecting a question in the question list, and the like.
  • Step S102 Determine a first answer corresponding to the question input by the user.
  • the customer service system may obtain a first answer corresponding to the entered question from a background database.
  • the first answer in this embodiment may be selected from a plurality of answers pre-built by a background database, or may be constructed in real time based on a question input by a user, which is not limited in this embodiment.
  • Step S103 Provide the first answer to the user.
  • the customer service system can push it to the electronic device used by the user to input the question for the user to learn.
  • the input mode may correspond to an answer different from the first answer, so that a suitable answer can be pushed to the user when a different question input mode may occur.
  • the question also corresponds to at least a second answer, and the second answer is an answer provided to the user in different question input modalities.
  • the problem input by the user in step S101 may be a problem that is frequently encountered or expected to be encountered in the customer service system.
  • this embodiment may correspond to multiple problems. Different answers, so that a suitable answer can be selected to be pushed to the user under different possible question input modalities, so that the customer service system implementing the answer providing method has a suitable answer provided under different possible question input modalities Ability.
  • a corresponding answer can be determined and provided based on a question input by a user, and can also provide different answers to the user in different question input modalities, thereby satisfying the user's different answers Demand to increase user satisfaction with answers.
  • the first answer and the second answer may be answers represented by different answer types, or answers having different contents, and thus may correspond to Input modalities for different problems that may occur in order to meet different user needs.
  • the first and second answers corresponding to the same question in different question input modes in the embodiments of the present disclosure are described in detail below.
  • the first answer and the second answer in the embodiments of the present disclosure may be answers represented by different answer types, where the answer types may include, but are not limited to, text types, video types, image types, and sound types.
  • a text type is a type that can present answers in the form of text on an electronic device that a user uses to enter a question.
  • the text can be text explaining a technical term (e.g., bootloader), indicating certain steps (e.g., setting VPN operation).
  • the video type is a type that can present answers on the electronic device in the form of a video (or more generally, in the form of pictures + text / speech).
  • the video can be a teaching video to show a technical term, a display Demonstration videos of certain steps, etc.
  • the image type is a type that can present answers in the form of pictures on the electronic device.
  • the picture may be a photo of a related topic that the user wants to retrieve, a picture showing the meaning of a technical term, and / or an indication of an operation step.
  • the sound type is a type in which answers can be presented in the form of voice on the electronic device.
  • the voice can be audio explaining a technical term, audio explaining some operation steps, and the like.
  • the first answer and the second answer in the embodiment of the present disclosure may have two different answer types selected from the above answer types, so as to meet different user needs.
  • the first answer and the second answer in the embodiments of the present disclosure may be answers having different contents.
  • the first answer and the second answer may be answers represented by the same answer type for the same question, but with different emphasis content to satisfy different user groups.
  • the first answer may be a more condensed version in which a part of the content is deleted from the answer content of the above-mentioned standard question and answer example
  • the second answer may be a more detailed version supplemented with a part of the content based on the answer content of the above-mentioned standard question and answer example, thereby Meet the needs of different user groups.
  • the first and second answers may correspond to the answers that the customer service system can provide in different question input modalities, and different question input modalities reflect different user needs.
  • the question input modal for the user to input the question may be determined based on various methods. The following describes several example ways to determine the question input modal.
  • questions can at least include mobile phones, laptops, tablets, smart speakers, personal digital assistants, and the like.
  • the question input modalities can be determined based on the device type and / or environmental parameters of the electronic device used when the user enters the question, so that different answers correspond to possible question input modalities.
  • the type of electronic device can indicate whether the electronic device supports display functions such as video / picture / text display, whether it supports the function of playing voice, etc., so it can be judged based on this whether the electronic device can support presentation of a certain type Answers to avoid users getting answers that their electronic devices cannot display / play and reduce user satisfaction. For example, when a user inputs a question through a smart speaker through voice, the type of the electronic device can be determined to be a smart speaker, which does not have a display screen but has the ability to play sounds.
  • the question input modal can be determined as An electronic device having only a sound playback function is used to input the modalities of the questions, and accordingly, the answers of the sound type can be corresponding to the answers that the customer service system can provide in the question input modalities.
  • the environmental parameters of the electronic device can indicate whether it is in the range covered by the wireless local area network, so the factors such as the bandwidth limitation and traffic cost of the answer transmission can be evaluated accordingly, so that the answer can be presented to the user in a cost-effective manner.
  • the user's electronic device is not connected to the wireless local area network, considering that the cost of the mobile phone's data is high, the user may not expect to spend too much data to get the answer.
  • the question can be entered into the modal Determined as a modal for inputting a question using an electronic device not connected to a wireless local area network, whereby a text-type answer with a small amount of data (for example, a text-type answer) can correspond to an answer that the customer service system can provide in the question input modal .
  • a text-type answer with a small amount of data for example, a text-type answer
  • the answer of the same type but more concise content may correspond to the answer that the customer service system can provide in the question input mode.
  • the electronic device is in the range covered by the wireless local area network as an example of its environmental parameters
  • other environmental parameters of the electronic device can also be considered, including but not limited to: the noise level around the electronic device, Location, electronic equipment operating status, etc.
  • the noise level is low and the user is located in a quiet place such as a library
  • the answer type without sound can correspond to the answer that the customer service system can provide in the question input mode; for example, when the electronic device is running navigation software
  • the answer of the sound type may correspond to the answer that the customer service system can provide in the question input mode, and so on.
  • the question input modalities can be determined based on the user's knowledge background and / or personal preferences, so that different answers correspond to possible question input modalities.
  • the identity of the user may be determined based on the identification information input by the user through the electronic device, and then the knowledge background of the user whose identity is determined may be determined.
  • the user's knowledge background can be the academic information, work experience, technical ability and other information provided by the user through a questionnaire in advance, and the user's ability to understand the answers to the questions can be estimated based on this information.
  • the question input modal can be determined as the modal of the technical person in the field to enter the question.
  • a simple type answer for example, a text type answer
  • And / or a more concise answer corresponds to the answer that the customer service system can provide in the question input mode, thereby saving the user's time cost and network overhead; if the user has less knowledge of the technical problem area, the question can be
  • the input modal is determined as the modal of the input question of the person in the non-relevant field.
  • a vivid answer for example, a video-type answer
  • / or a more detailed answer may correspond to the ability of the customer service system to The answer provided in the question input mode, so that the user can fully understand the answer content of the question.
  • the identity of the user can be determined, and then the personal preferences of the user whose identity is determined can be determined based on which different question input modalities can be determined.
  • the user's personal preferences may be preference information set by the user on his electronic device, from which different question input modalities may be determined.
  • the preference information may be information provided by the user through a questionnaire in advance that they are more inclined to view a certain type of answer, whether they have sufficient time to learn knowledge, and whether they are particularly concerned about a certain field of problems.
  • the user's personal preferences can indicate that the user likes to get answers in a more vivid way of answering questions such as voice / video.
  • the question input modalities can be determined as interactive user input question modalities.
  • the Types of answers such as vivid voice / video correspond to the answers that the customer service system can provide in this question input mode.
  • the various questions entered by the user may correspond to the answer content of different structures (for example, compliant structure, non-compliant structure, etc.), and the answer content of different structures may be suitable to be represented by different types of answers, respectively.
  • the question input modalities can be determined based on the analysis of the answer content of the question input by the user, so that different types of answers correspond to the possible question input modalities.
  • the answer content in the above examples 1 and 2 contains more operation steps, which indicates that the answer content is an operation type, and the question input modal can be determined as the input operation related question.
  • the answer of the video type can correspond to the answer that the customer service system can provide in the question input modal, thereby helping the user to intuitively understand and grasp the sequence of each operation step.
  • the answer content of the above example 3-5 is an explanation of definition and / or clarification of facts, which indicates that the content of the answer is definition / factual content, and the question input modal can be determined as input non-operation related questions.
  • the text format may be sufficient for the user to understand, so the answer of the text type may correspond to the answer that the customer service system can provide in the question input modal.
  • the method of determining the answer content as operation / definition / fact based on analysis of the answer content will be described below.
  • the above describes an example of determining the question input modalities based on factors such as the device type and / or environmental parameters of the electronic device used by the user to input the question, the user's knowledge background and / or personal preference, the answer content of the question, and other factors.
  • factors such as the device type and / or environmental parameters of the electronic device used by the user to input the question, the user's knowledge background and / or personal preference, the answer content of the question, and other factors.
  • the question input modalities can also be determined based on other factors, and according to this, different answers can be corresponding to different question input modalities that may appear for pushing to users with different needs. .
  • this embodiment may receive a question input by a user, determine a first answer corresponding to the question input by the user, and provide the first answer to the user, thereby The user can be made to learn through the first answer to obtain knowledge.
  • the first answer corresponding to the question input by the user can be determined in various ways. In the following, the question determined in step S102 of FIG. 1 and determined by the user will be described with reference to FIGS. 2A and 2B. Several exemplary methods corresponding to the first answer.
  • the method includes the following steps:
  • step S201 multiple answers corresponding to the questions input by the user are obtained.
  • the first answer may be selected from a plurality of pre-structured answers, in which case, a plurality of pre-structured answers corresponding to the user-entered question may be obtained first for subsequent screening to find suitable The first answer pushed to the user.
  • the multiple answers in this embodiment may be multiple answers that are pre-built and represented by different answer types, or multiple answers with different contents.
  • Step S203 Determine the first answer from the multiple answers.
  • the first answer may be determined from the obtained multiple answers according to various criteria. As a possible example, similar to what was previously described on how to determine the problem input modalities, it may be based on the device type and / or environmental parameters of the electronic device the user uses to input the problem, the user ’s knowledge background and / or personal preferences, The content of the answer to the question and other factors determine the first answer from the multiple answers obtained.
  • the method includes the following steps:
  • Step S202 Determine an answer type used to represent the answer. It can be understood that the first answer determined according to the question input by the user in step S101 above is to be presented to the user in a certain type, and selecting an appropriate type of answer can satisfy the user's needs. Similar to the description of determining the first answer from the multiple answers obtained in step S203, based on the device type and / or environmental parameters of the electronic device used by the user to input the question, the user ’s knowledge background and / or personal preference, The answer content of the question and other factors determine the answer type used to represent the answer, so that subsequent types of answers can be obtained from the background server.
  • an answer represented by the determined answer type is obtained as the first answer.
  • a first answer of a determined answer type may be selected from a plurality of answers pre-built by the background database to provide it to the user; for example, when the plurality of answers pre-built by the background database does not include any certainty
  • the answer represented by the determined answer type may be constructed in real time and provided to the user, and the embodiments of the present disclosure are not limited thereto.
  • the above describes several exemplary methods of determining the first answer corresponding to the question input by the user in conjunction with FIG. 2A and FIG. 2B, and described in step S202 of FIG. 2B that an answer representing the answer can be determined based on various factors Type for subsequent push to the user to get answers of that type.
  • an exemplary method of determining an answer type used to represent an answer based on the answer content of the question in step S202 of FIG. 2B will be described in detail with reference to FIG. 3.
  • the method in this embodiment may include the following steps:
  • Step S301 Perform feature extraction on the answer content of the answer. It can be understood that the answer content contains standard content that can effectively answer the user's questions, and it needs to be expressed in a certain type for pushing to the user.
  • the extracted features include at least one of an inherited symbol feature, an inherited word feature, an interop word feature, and an interop sequence length feature.
  • the inheritance symbol feature may be the number of occurrences of the inheritance symbol "->" as in the example 1 of the standard question and answer, so as to represent the successive inheritance relationship of each operation step.
  • the inheritance word feature may be, for example, the inheritance word frequency to characterize the frequency of occurrence of the inheritance word such as "first", "then", and the like in the standard question answering example 2.
  • Interaction word characteristics may be related to the frequency of operation-type words in system software and application software to characterize the frequency of occurrence of inter-operation-type words such as "settings", "VPN", "more” in the standard question answering examples 1 and 2.
  • the length of the interactive operation sequence may be the number of consecutive occurrences of operation words in the system software and application software to characterize the total length and complexity of the interactive operation sequence. It can be understood that the features listed above are only examples, and different feature extraction methods can also be used to extract different kinds of features that can reflect the content of the answer according to actual needs.
  • Step S302 Determine the structure of the answer content according to the extracted features.
  • the structure of the answer content may be determined as an inherited structure or a non-adherent structure.
  • the answer content of the standard question answer examples 1 and 2 may be determined to have an inherited structure
  • the answer content of the question answer examples 3-5 may be determined to have a non-adherent structure.
  • Step S303 Based on the structure of the answer content, determine each priority score of the answer in a plurality of different answer types. For example, considering that the answer content with the inheritance structure contains more operation steps, it is more helpful for users to understand it by the video type. Therefore, for the answer content with the inheritance structure, a higher priority score can be given to the video type. The other types are given lower priority scores. As another example, considering that the content of the non-adherent structure is mostly explanatory or descriptive description, it is sufficient for the user to understand and save the user ’s browsing time and transmission bandwidth. Therefore, The content of the answer can be given a higher priority score for the text type, and a lower priority score for other types.
  • the priority score may be a priority score value (for example, a score value in a range of 0 to 100), or an order of priorities (for example, first priority, second priority, etc.) ).
  • the video type of the answer content of the inheritance structure can be set to the first priority, and the remaining types are set to the second priority; the text type of the answer content of the non-heritage structure can be set to the first priority. Level, and set the remaining types as the second priority.
  • Step S304 Determine an answer type for representing an answer based on the priority scores.
  • the type of answer suitable for pushing to the user can be determined, so that the answer content is finally expressed in this type and pushed to the user, making the answer easier to be User understands and accepts.
  • the answer may be represented by the type with the highest priority score to present to the user for learning.
  • the answer providing method of the embodiment of the present disclosure it is possible to determine the respective priority scores representing answers in different answer types by analyzing the content of the answers, and automatically determine the answer type that is most suitable for pushing to the user accordingly, thereby avoiding a lot of manual Annotate the problems caused by waste of resources and low efficiency, and enable users to better understand the content of answers, and improve user satisfaction with answers.
  • each priority score representing an answer with different answer types based on the content of the answer, and after determining each priority score, it is expected that the answer type with the highest priority score represents the answer.
  • the following situations may exist: for the questions entered by the user, several types of answers have been constructed in advance, but the answers with the highest priority scores have not yet been constructed; or, the questions entered by the user are new in this field The question that has arisen has not yet been constructed with any type of answer. In this case, because the answer has not yet been constructed with the type with the highest priority score, it is temporarily impossible to represent the answer with that type and push it to the user.
  • the method for determining each priority score in this embodiment may include the following steps:
  • step S401 it is confirmed whether an answer represented by an answer type having the highest priority score is constructed in advance.
  • step S402 when an answer represented by the answer type with the highest priority score is constructed in advance, it is confirmed that the answer is represented by the answer type with the highest priority score.
  • the answer type with the highest priority score has been constructed in advance, and the customer service system can obtain this type of answer as the first answer to provide the user with learning.
  • step S403 when the answer represented by the answer type with the highest priority score has not been constructed, it is determined whether to construct the answer represented by the answer type with the highest priority score.
  • the answer type with the highest priority score has not yet been constructed, so the customer service system can consider various factors to decide whether to construct the answer of the answer type with the highest priority score. If it is decided to construct the answer represented by the answer type with the highest priority score, the customer service system can provide the constructed answer to the user as the first answer. If you decide not to construct the answer represented by the answer type with the highest priority score, you can further decide whether to construct the answer represented by the answer type with the second priority score, and decide whether to construct the answer represented by the answer type with the third priority score. , And so on, so as to finally determine the type used to represent the answer, and then the constructed answer can be provided to the user as the first answer.
  • the cost of constructing an answer with a video type is relatively high
  • the cost of constructing an answer with a text type is relatively low
  • the cost of constructing an answer with a voice and picture type is between the two.
  • high-frequency questions may be more worthwhile to spend resources to build higher-cost types of answers
  • low-frequency questions spend higher costs to construct answers will cause waste of resources. Therefore, it is important to balance the cost of construction and the usage of answers when constructing the answer.
  • various factors may be considered to decide whether to construct the answer of the answer type with the highest priority.
  • the decision of whether to construct the highest priority score in step S403 of FIG. 4 will be described with reference to FIG. 5.
  • the answer type represents an exemplary implementation of the answer. As shown in FIG. 5, the method may include the following steps:
  • Step S501 calculating a construction cost for constructing the answer with the highest priority answer type.
  • the calculation method of the construction cost is different.
  • the answer of the inherited structure and the answer of the non-adhered structure are taken as examples to describe the construction cost of the answer.
  • calculating the construction cost of constructing the answer with the highest priority answer type may include the following steps:
  • the inheritance structure answer is an answer that includes multiple operation steps with inheritance relationships.
  • the above considerations are for the calculation of the step construction cost for each operation step. In fact, if some of the operation steps have been prepared in advance without repeating the construction, the steps of those operation steps can be constructed. The cost is set to zero or a relatively low value to accurately reflect the cost of answer production in the actual scenario.
  • the step construction cost for constructing the above 4 operation steps with the answer type (eg, video type) with the highest priority score can be calculated separately, and then the construction cost for constructing the entire answer with this type is calculated.
  • the pre-built step cost knowledge base may be constructed based on the ease of operation steps.
  • the pre-built step cost knowledge base involves most of the operational steps that can be decomposed in the field of mobile phones, such as "enter settings", "open camera”, "about phone” and other general operating steps.
  • a single operation step is a unit that manages the step construction cost of building the single step in various different types (such as video, text, etc.). Therefore, according to the keywords of each operation step to be retrieved, the step cost knowledge base can be retrieved to obtain each step construction cost corresponding to each operation step for subsequent calculation of the construction cost of the entire answer.
  • the method may include the following steps:
  • step S601 the step construction cost of each operation step of constructing the answer content of at least one obedient structure with at least one answer type is determined in advance.
  • the at least one answer type described above covers at least the answer type with the highest priority score to be calculated, and the answers of the at least one obedient structure described above include a plurality of corresponding ones such as common operational interaction questions in the mobile phone field. Support structure answer.
  • each atomized operation step can be obtained for the step cost knowledge base to manage the construction cost in steps.
  • the step cost knowledge base can cover different types of answer types, and during its construction process, different numbers of answers from the inherited structure in the field can be collected to construct step cost knowledge bases with different levels of knowledge reserve.
  • the material for constructing the video answer can be obtained simply by taking screenshots or recording screens of the software operations, etc., which can be combined on the basis of the benchmark cost. Relatively low cost.
  • operation steps other than system software and application software for example, when it is relatively difficult to obtain a video production material by connecting a laptop computer for operation, a relatively high cost can be combined on the basis of the baseline cost.
  • the method of calculating the cost described above is only an exemplary description, and those skilled in the art may assign different operation steps to different construction steps according to actual conditions and production capabilities of the production personnel.
  • Step S602 Construct a set of operation steps based on each operation step of the answer of the at least one obedient structure.
  • the set of operation steps covers the common operation steps in the field of mobile phones.
  • the operation steps corresponding to the operation steps s 1 to s 4 in the standard question answering example 1 can be found from the set SS.
  • the number of answers of the obedient structure collected is larger, the larger the range of operation steps covered by the operation step set is, the higher the level of the knowledge reserve level of this step in the knowledge base.
  • step S603 a step cost knowledge base is constructed by using a single operation step in the operation step set as a unit.
  • the constructed step cost knowledge base is shown in FIG. 6B, where each operation step ss i in the set SS of operation steps has a corresponding cost Css i for constructing an answer in one or more types.
  • the inventors of the present disclosure recognize that in the set of steps, the importance of each operation step is not the same, and the importance of the step is also an important factor to measure the priority of the step. For example, if the importance of an operation step is high, it means that the operation step is more common and necessary in the operation of the mobile phone field. Therefore, the material that constructs the operation step is used many times in the construction of other answers. To avoid wasting resources caused by repeated production of individual operating steps. In other words, the higher the importance of an operation step, it means that it is more worthwhile to construct the answer of the operation step.
  • the cost knowledge base constructed in step S603 described above includes a plurality of atomized operation steps ss 1 to ss N.
  • Each atomicization step ss 1 ⁇ ss N has two kinds of inheritance relationship of pre-order / post-sequence in the collected original inheritance structure answer.
  • the pre- / post-order relationship in can be used to construct a step-directed graph with all atomic steps as nodes and pre- / post-order relationships as edges. Further, based on the step directed graph, the importance of a single atomization step can be calculated. For example, as an exemplary method of calculating the importance of each step, a PageRank algorithm can be used for implementation.
  • the step ss 1 ⁇ ss 3 as an example, the respective steps ss 1 ⁇ ss 3 has a preamble / rear relation between the order and the other steps, such as subsequent ss ss. 3 of 2, 2 is ss ss Follow-up of 1 .
  • FIG. 6C schematically illustrates that there is a pre-order / post-order relationship between each step and another step, those skilled in the art can understand that since each step may originate from the different collected
  • the answer of the original inheritance structure of the structure is that each step can have a pre-order / post-order relationship with two or more steps, which is not limited in the present invention.
  • step construction cost of each operation step in the step cost knowledge base can be weighted, and the step cost knowledge base can be updated with the weighted step construction cost.
  • the weighted cost of step ss i can be Css i / SRss i .
  • An inherited structure answer refers to an answer that does not include multiple operation steps with an inheritance relationship, and examples thereof may include: an operation-type answer without a inheritance structure, a fact-type answer, and a definition-type answer.
  • an operation-type answer without a inheritance structure may include: an operation-type answer without a inheritance structure, a fact-type answer, and a definition-type answer.
  • different methods can be used to calculate the construction cost of their answers. An exemplary method of calculating the construction cost of a non-adherent structure answer is described below.
  • calculating the construction cost of the answer type with the highest priority score may include the following steps:
  • (A) Determine whether the answer content is an operation content.
  • a feature extraction method may be used to extract the inter-operation word features in the answer, and based on the extracted features combined with semantic analysis to determine whether the answer is an operation-type answer.
  • the standard question answering example 3 it is possible to extract the features of the interactive operation words such as "download”, and determine the content of the answer as an operation-type answer in combination with context semantic analysis.
  • step (B) When it is determined in step (A) that the answer content is operation-type content, the construction cost is calculated based on the degree of ease of operation corresponding to the operation-type content. For example, similar to that described above in connection with the construction of an inherited answer cost, the baseline cost of constructing the answer with the highest priority answer type can be determined, and further combining the ease of operation of the steps in the answer, determine the The answer type with the highest priority score is the cost of constructing that answer.
  • step (C) When it is determined in step (A) that the answer content is not an operation content, it is further determined whether the answer content is a fact content or a definition content.
  • a feature extraction method may be used to extract features in the answer, etc., and determine whether the answer is fact-based content or definition-based content based on the extracted features. For example, through the feature extraction and recognition method, it can be determined that the answer in Example 4 is definition-type content, and the answer in Example 5 is fact-type content.
  • step (D) When it is determined in step (C) that the answer content is fact-type content or definition-type content, calculate the construction cost based on the complexity of the answer content and the technical background of the relevant user. For example, some technical terms or definitions are relatively complicated or are relatively unfamiliar to a large user group. A large number of explanations are required to enable users to grasp their meanings. Therefore, more production materials need to be prepared, which can set higher construction costs.
  • Example 4 of the standard question and answer above While some answers are relatively simple, you can set a lower construction cost for them, such as the standard question and answer example 5 above.
  • Step S502 predicting a usage rate of the answer.
  • the exemplary method for predicting the usage rate of answers may include the following steps:
  • a method for classifying multiple historical problems of a user may use a traditional classifier such as an SVM and a neural network classification model CNN, and the embodiments of the present disclosure are not limited thereto.
  • SVM SVM
  • CNN neural network classification model
  • a pre-trained word vector can be used as an initial value to vectorize the user's historical problem and standard problem, and then use algorithms such as CNN, LSTM, etc.
  • the historical problem and the standard problem are encoded to obtain a sentence vector, and according to the sentence vector, it is determined whether the user's historical problem and the standard problem match.
  • Figure 7 shows the matching results of the user's historical and standard questions, where historical question q1 matches the standard question Q2, historical question q2 matches the standard question Q1, historical question q3 matches the standard question Q2, and historical question q4 Matches standard question Q3, and historical question q5 matches standard question Q3.
  • (C) Predicting the usage rate of the answer based on the number of historical questions matching the standard question among the multiple historical questions.
  • the standard question corresponds to the answer to be calculated for its usage rate.
  • the number of times that the standard question is asked can be determined, and the usage rate of the answer is reflected accordingly. For example, in combination with the matching results of the historical questions and standard questions of the user in FIG. 7, it is determined that the number of historical questions matching the standard question Q1 is 1, the number of historical questions matching the standard question Q2 is 2, and the standard question Q3 is matched.
  • the number of historical questions is 2, and the frequency of using the standard answers A1 to A3 can be estimated according to the correspondence between the user's historical questions, standard questions, and answers whose usage rate is to be calculated. For example, the ratio of the number of user historical questions corresponding to the answer A1 to the total number of historical questions may be used as the usage rate of the answer A1.
  • the usage rate of the answer is predicted based on the number of historical questions. Understandably, other methods can be used for prediction. For example, when a question and its answer are relatively new in the field, the number of historical questions corresponding to the answer may be relatively limited. In this case, it can be based on the hot trend of the question, the group of interest Example of the number and other factors to estimate the usage rate of the answer.
  • the method further includes: Step S503, based on the construction cost and usage rate, decide whether to construct an answer represented by an answer type with the highest priority score. Assuming that the construction cost of the above answer A is C (A) and the utilization rate is U (A), the ratio U (A) / C (A) between the two can be calculated. Further, the ratio is compared with a predetermined threshold value. When the ratio is greater than the predetermined threshold value, it indicates that the answer type with the highest priority score is worth representing the answer, and it may be decided to construct the answer type with the highest priority score.
  • the predetermined threshold may be set according to user requirements, and the present invention is not limited thereto.
  • the user's overall satisfaction evaluation of the answer content is also an important factor in measuring the answer.
  • the above ratio may be weighted according to the average user satisfaction score, and the weighted ratio may be used to decide whether to construct an answer represented by the answer type with the highest priority score. For example, assuming that the user ’s overall satisfaction with the answer content is SF (A), the above-mentioned ratio U (A) / C (A) may be multiplied by the weighting factor SF (A), and then whether to construct a priority score The answer represented by the highest answer type.
  • each priority score representing the answer in different answer types is determined, and the answer type suitable for being pushed to the user is automatically determined according to this, which can avoid a lot of manual annotation and allow the user to better understand the answer.
  • Content to increase user satisfaction with answers when deciding whether to construct an answer of a certain type, the construction cost of the type of answer and its usage rate can be considered, thereby taking into account the operating cost of the customer service system and user satisfaction.
  • FIG. 8 shows an exemplary structural block diagram of an answer providing apparatus 800 according to an embodiment of the present disclosure.
  • the answer providing device 800 may be any type of electronic device such as a laptop computer, a server, a cloud data device, and the like.
  • the answer providing device 800 may receive a question input by a user and provide a corresponding answer to the user. In the following, only the main functions of the units of the answer providing device 800 are described, and details that have been described above are omitted.
  • the answer providing device 800 may include a question receiving unit 801, an answer determining unit 802, and an answer providing unit 803.
  • the answer providing device 800 may further include other components not shown in FIG. 8, such as a display, an input / output device, and the like.
  • the question receiving unit 801 may receive a question input by a user.
  • the user may use the electronic device that he holds or the answer providing device itself to input various questions to the customer service system in various ways.
  • the manner in which the user enters the question may include, but is not limited to: typing input, voice input, selecting a question in the question list, and the like.
  • the answer determination unit 802 may determine a first answer corresponding to a question input by the user. For example, after receiving a question entered by a user, the customer service system may obtain a first answer corresponding to the entered question from a background database. It should be noted that the first answer in this embodiment may be selected from a plurality of answers that are pre-built by a background database, or may be constructed in real time based on a question input by a user, which is not used in the embodiments of the present disclosure. limit.
  • the answer providing unit 803 may provide the first answer to the user. For example, after determining the first answer, the customer service system can push it to the electronic device that the user uses to enter the question. It should be noted that although it is described that the answer providing unit 803 provides the first answer to the user, this embodiment also considers different user needs in different question input modalities, so the questions input by the user in this embodiment are different.
  • the question input modal may correspond to an answer different from the first answer, so that the answer providing unit 803 can push a suitable answer to the user when a different question input modal may occur.
  • the question also corresponds to at least a second answer, and the second answer is an answer provided to the user in different question input modalities.
  • the question entered by the user may be a question frequently encountered or expected to be encountered in the customer service system.
  • this embodiment may correspond to the question with multiple different answers in order to
  • the answer providing unit 803 can select a suitable answer to be pushed to the user in different possible question input modalities, so that the customer service system executing the answer providing method has a suitable answer in different possible question input modalities.
  • the answer providing device 800 may also perform any processing in the answer providing method described previously in conjunction with FIG. 1 to FIG. 7, and specific details are not described herein.
  • At least some of the modules, sub-modules, units, or sub-units according to the embodiments of the present disclosure, or at least part of the functions of any of them, can be implemented in one module. Any one or more of the modules, sub-modules, units, and sub-units according to the embodiments of the present disclosure may be implemented by being split into multiple modules.
  • any one or more of the modules, sub-modules, units, and sub-units according to the embodiments of the present disclosure may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), System-on-chip, system-on-substrate, system-on-package, application-specific integrated circuit (ASIC), or any other reasonable means of hardware or firmware that can integrate or package the circuit, or software, hardware, and firmware Any one of the implementation manners or an appropriate combination of any of them may be implemented.
  • FPGA field programmable gate array
  • PLA programmable logic array
  • ASIC application-specific integrated circuit
  • any one of the implementation manners or an appropriate combination of any of them may be implemented.
  • one or more of the modules, submodules, units, and subunits according to the embodiments of the present disclosure may be at least partially implemented as computer program modules, and when the computer program modules are executed, corresponding functions may be performed.
  • any number of the question receiving unit 801, the answer determining unit 802, and the answer providing unit 803 may be combined into one module for implementation, or any one of the modules may be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules may be combined with at least part of the functions of other modules and implemented in one module.
  • At least one of the question receiving unit 801, the answer determining unit 802, and the answer providing unit 803 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array ( PLA), system-on-chip, system-on-substrate, system-on-package, application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging the circuit, such as hardware or firmware, or software or hardware And any one of the three implementations of firmware or any appropriate combination of them.
  • FPGA field programmable gate array
  • PLA programmable logic array
  • ASIC application-specific integrated circuit
  • at least one of the question receiving unit 801, the answer determining unit 802, and the answer providing unit 803 may be at least partially implemented as a computer program module, and when the computer program module is executed, a corresponding function may be performed.
  • the answer providing device can determine and provide corresponding answers based on a question input by a user, and can also input modalities in different questions Provide users with different answers to meet the different needs of users and improve user satisfaction with answers.
  • each priority score representing the answer in different answer types is determined, and the answer type suitable for being pushed to the user is automatically determined according to this, which can avoid a lot of manual annotation and allow the user to better understand the answer.
  • Content to increase user satisfaction with answers.
  • the construction cost of the type of answer and its usage rate can be considered, thereby taking into account the operating cost of the customer service system and user satisfaction.
  • FIG. 1 Another embodiment of the present disclosure also provides a computer-readable storage medium for storing non-transitory computer-readable instructions, which can be executed when the non-transitory computer-readable instructions are executed by a computer.
  • the answer provides the method.
  • the computer-readable storage medium may be included in the device / apparatus / system described in the above embodiments; it may also exist alone without being assembled into the device / apparatus / system.
  • the foregoing storage medium includes, but is not limited to, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory ( CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the foregoing.
  • a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in combination with an instruction execution system, apparatus, or device.
  • each block in the flowchart or block diagram may represent a module, program segment, or part of code, which contains one or more of the logic functions used to implement the specified logic. Executable instructions.
  • the functions labeled in the blocks may also occur in a different order than those labeled in the drawings. For example, two blocks represented one after the other may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved.
  • each block in the block diagram or flowchart, and combinations of blocks in the block diagram or flowchart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with A combination of dedicated hardware and computer instructions.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Electrically Operated Instructional Devices (AREA)

Abstract

一种答案提供方法和设备。所述答案提供方法包括：接收用户输入的问题（S101）；确定与用户输入的问题相对应的第一答案（S102）；以及将所述第一答案提供给用户，其中，所述问题至少还对应于第二答案，所述第二答案是在不同的问题输入模态下提供给用户的答案（S103）。根据该答案提供方法，可以基于用户输入的问题确定并且提供相对应的答案，并且还能够在不同的问题输入模态下给用户提供不同的答案，由此满足用户对答案的不同需求，提高用户对答案的满意度。

Description

一种答案提供方法及设备 技术领域
本公开涉及电子设备的领域，更具体地，本公开涉及一种答案提供方法及设备。
背景技术
如今，诸如客服系统等问答系统的使用越来越普遍，用户可以输入想要咨询的问题，然后得到客服系统推送的答案，以满足用户对各种信息的需求。
在目前的客服系统中，针对用户可能会提出的问题，需要领域专家人工判断与之对应的答案内容来预先设定唯一答案。进而，当用户输入该问题时，客服系统便会向用户推送预先设定好的唯一答案。然而，预先设定好的唯一答案的内容固定并且类型单一，无法满足用户的不同需求，导致用户对客服系统推送答案的满意度降低。另一方面，预先设定唯一答案的过程需要领域专家对答案的类型做出人工判断，花费了大量的人工标注成本。
发明内容
为了解决现有技术中的上述技术问题，根据本公开的一个方面，提供一种答案提供方法，包括：接收用户输入的问题；确定与用户输入的问题相对应的第一答案；以及将所述第一答案提供给用户，其中，所述问题至少还对应于第二答案，所述第二答案是在不同的问题输入模态下提供给用户的答案。
根据本公开的另一个方面，提供一种答案提供设备，包括：问题接收单元，配置为接收用户输入的问题；答案确定单元，配置为确定与用户输入的问题相对应的第一答案；以及答案提供单元，配置为将所述第一答案提供给用户，其中，所述问题至少还对应于第二答案，所述第二答案是在不同的问题输入模态下提供给用户的答案。
根据本公开的另一个方面，提供一种计算机可读存储介质，用于存储非暂时性计算机可读指令，当所述非暂时性计算机可读指令由计算机执行时可以执行本公开的实施例的答案提供方法。
根据本公开的另一个方面，提供一种计算机程序，所述计算机程序包括计算机可执行指令，当所述计算机可执行指令被执行时可以执行本公开的实施例的答案提供方法。
根据本公开各个方面的答案提供方法及设备，可以基于用户输入的问题确定相对应的答案并且能够在不同的模态下给用户提供不同的答案，由此满足用户对答案的各种不同需求，提高用户对答案的满意度。
附图说明
从下面结合附图对本公开实施例的详细描述中，本公开的这些和/或其它方面和优点将变得更加清楚并更容易理解，其中：
图1是图解根据本公开实施例的答案提供方法的流程图；
图2A和图2B是图解根据本公开实施例的确定与用户输入的问题相对应的第一答案的示例性方法；
图3是图解根据本公开实施例的基于问题的答案内容确定用来表示答案的答案类型的示例性方法；
图4是图解根据本公开实施例的基于所述各个优先级得分确定用来表示答案的答案类型的示例性方法；
图5是图解根据本公开实施例的决定是否要构建以优先级得分最高的答案类型表示的答案的示例性方法；
图6A是图解根据本公开实施例的构建步骤成本知识库的示例性方法的流程图；
图6B是图解根据本公开实施例的所构建的步骤成本知识库的示意图；
图6C是图解根据本公开实施例的对步骤成本知识库中的各个操作步骤的步骤构建成本加权的示意图；
图7是图解根据本公开实施例的用户的历史问题和标准问题的进行匹配的示意图；
图8是图解根据本公开实施例的答案提供设备的框图。
具体实施方式
以下，将参照附图来描述本公开的实施例。但是应该理解，这些描述只是示例性的，而并非要限制本公开的范围。在下面的详细描述中，为便于解释，阐述了许多具体的细节以提供对本公开实施例的全面理解。然而，明显地，一个或多个实施例在没有这些具体细节的情况下也可以被实施。此外，在以下说明中，省略了对公知结构和技术的描述，以避免不必要地混淆本公开的概念。
在此使用的术语仅仅是为了描述具体实施例，而并非意在限制本公开。在此使用的术语“包括”、“包含”等表明了所述特征、步骤、操作和/或部件的存在，但是并不排除存在或添加一个或多个其他特征、步骤、操作或部件。
在此使用的所有术语(包括技术和科学术语)具有本领域技术人员通常所理解的含义，除非另外定义。应注意，这里使用的术语应解释为具有与本说明书的上下文相一致的含义， 而不应以理想化或过于刻板的方式来解释。
在使用类似于“A、B和C等中至少一个”这样的表述的情况下，一般来说应该按照本领域技术人员通常理解该表述的含义来予以解释(例如，“具有A、B和C中至少一个的系统”应包括但不限于单独具有A、单独具有B、单独具有C、具有A和B、具有A和C、具有B和C、和/或具有A、B、C的系统等)。在使用类似于“A、B或C等中至少一个”这样的表述的情况下，一般来说应该按照本领域技术人员通常理解该表述的含义来予以解释(例如，“具有A、B或C中至少一个的系统”应包括但不限于单独具有A、单独具有B、单独具有C、具有A和B、具有A和C、具有B和C、和/或具有A、B、C的系统等)。
附图中示出了一些方框图和/或流程图。应理解，方框图和/或流程图中的一些方框或其组合可以由计算机程序指令来实现。这些计算机程序指令可以提供给通用计算机、专用计算机或其他可编程数据处理装置的处理器，从而这些指令在由该处理器执行时可以创建用于实现这些方框图和/或流程图中所说明的功能/操作的装置。本公开的技术可以硬件和/或软件(包括固件、微代码等)的形式来实现。另外，本公开的技术可以采取存储有指令的计算机可读存储介质上的计算机程序产品的形式，该计算机程序产品可供指令执行系统使用或者结合指令执行系统使用。
在本公开的以下实施例中，以手机领域中的客服系统作为问答系统的示例描述了答案提供方法及设备。本领域技术人员可以理解，本公开的答案提供方法及设备同样可以应用于其他领域的问答系统中。为了便于说明和理解，以下列举手机领域中的几个标准问答的内容示例作为参考。
<标准问答示例1>
标准问题：如何找到手机的IMEI？
标准答案：进入设置->关于手机->状态->IMEI信息。
<标准问答示例2>
标准问题：如何设置VPN？
标准答案：首先，进入设置，然后按下“更多”按钮，您会在右上角找到“VPN”。如果您不了解设置VPN所需的信息，请联系VPN提供商。
<标准问答示例3>
标准问题：在哪里获得PC或Mac的USB驱动程序？
标准答案：您可以从以下链接下载官方USB驱动程序：
Windows：http://www.motorola.com/getmdmwin
Mac OS X：http://www.motorola.com/getmdmmac
<标准问答示例4>
标准问题：解锁bootloader是什么意思？
标准答案：如果开发者或用户想要进一步访问系统文件或者运行定制固件，则可以解锁bootloader。Bootloader是设备启动时验证软件是否是官方时运行的一段代码。除非您有这方面的经验，否则我们不建议解锁bootloader。
<标准问答示例5>
标准问题：XX手机的显示屏尺寸多大？
标准答案：5.5英寸。
如先前所讨论，在目前的客服系统中，向用户推送的答案内容固定且类型单一。例如，对于每个上述标准问答，现有的手机客服系统仅提供以单一类型表示且内容固定的唯一答案，以向所有咨询该问题的用户均推送该唯一答案。本发明人认识到，不同的问题输入模态对应于不同的用户需求，例如，某一用户输入模态可能反映出用户仅支持聆听语音类型的答案来学习某些操作方法(例如，学习设置VPN的操作方法)，但现有的客服系统可能仅推送视频类型的答案；又例如，另一用户输入模态可能反映出用户希望获得技术术语更加精简/详尽的解答(例如，掌握bootloader的含义)，但现有的客服系统向所有用户推送详细程度完全相同的答案。因此，现有客服系统向用户推送的唯一答案可能对于不同用户和/或同一用户的不同需求来说并不合适，无法很好地满足不同问题输入模态下的用户需求。
有鉴于此，本公开的实施例提出一种答案提供方法，基于用户输入的问题确定并且提供与之相对应的答案，并且能够在不同的问题输入模态下给用户提供不同的答案，由此向用户推送适合的答案。下面，将参照图1描述根据本公开实施例的答案提供方法。如图1所示，该实施例的答案提供方法可以包括如下步骤：
步骤S101，接收用户输入的问题。例如，用户可以使用其持有的电子设备通过各种方式向客服系统输入想要询问的问题。用户输入问题的方式可以包括但不限于：打字输入、语音输入、在问题列表中选择问题等。
步骤S102，确定与用户输入的问题相对应的第一答案。例如，在接收到用户输入的问题后，客服系统可以从后台数据库获得与所输入的问题相对应的第一答案。需说明的是，本实施例中的第一答案可以从是后台数据库预先构建的多个答案中选择的，或者是基于用户所输入的问题而实时构建的，本实施例不以此为限。
步骤S103，将所述第一答案提供给用户。例如，在确定了第一答案之后，客服系统便可将其推送给用户用来输入问题的电子设备让用户学习。需说明的是，虽然该步骤描述了将第一答案提供给用户，但本实施例还考虑了不同的问题输入模态下的不同用户需求，因此本实 施例中用户输入的问题在不同的问题输入模态下可以对应于与第一答案不同的答案，由此便能够在可能出现不同的问题输入模态时向用户推送合适的答案。在本实施例中，所述问题至少还对应于第二答案，所述第二答案是在不同的问题输入模态下提供给用户的答案。例如，上述步骤S101中用户输入的问题可能是客服系统中经常遇到或者预计将要遇到的问题，考虑不同的问题输入模态下不同的用户需求，本实施例可以将该问题对应于多个不同答案，以便能够在可能出现的不同的问题输入模态下选择合适的答案来向用户推送，使得执行该答案提供方法的客服系统具有在可能出现的不同的问题输入模态下提供合适的答案的能力。
根据本公开实施例的答案提供方法，可以基于用户输入的问题确定并且提供相对应的答案，并且还能够在不同的问题输入模态下给用户提供不同的答案，由此满足用户对答案的不同需求，提高用户对答案的满意度。
以上结合图1描述了本公开实施例的答案提供方法，在该实施例中，第一答案和第二答案可以是以不同答案类型表示的答案，或者是具有不同内容的答案，由此可以对应于可能出现的不同的问题输入模态，以便满足不同的用户需求。以下具体描述本公开实施例中不同的问题输入模态下对应于同一问题的第一答案和第二答案。
一方面，本公开实施例中的第一答案和第二答案可以是以不同答案类型表示的答案，其中答案类型可以包括但不限于：文本类型、视频类型、图像类型和声音类型。文本类型是可以在用户用来输入问题的电子设备上以文字形式呈现答案的类型，例如，该文字可以是解释某一技术术语(例如，bootloader)的文字、指示某些操作步骤(例如，设置VPN的操作)的指示性文字等。视频类型是可以在该电子设备上以视频形式(或者更一般地，以图片+文字/语音解说形式)呈现答案的类型，例如，该视频可以是用于讲解某一技术术语的教学视频、展示某些操作步骤的演示视频等。图像类型是可以在电子设备上以图片形式呈现答案的类型，例如，该图片可以是用户想要检索的相关主题的照片、展示技术术语含义和/或操作步骤指示的图片等。声音类型是可以在电子设备上以语音形式呈现答案的类型，例如，该语音可以是讲解某一技术术语的音频、讲解某些操作步骤的音频等。本公开实施例中的第一答案和第二答案可以具有从以上答案类型中选择的两种不同的答案类型，从而满足不同的用户需求。
另一方面，本公开实施例中的第一答案和第二答案可以是具有不同内容的答案。例如，第一答案和第二答案可以是针对同一问题以同一种答案类型表示的答案，但是具有不同的侧重点的内容以满足不同用户群体。比如，第一答案可以是从上述标准问答示例的答案内容中删减一部分内容的更精简版本，而第二答案可以是在上述标准问答示例的答案内容基础上补充一部分内容的更详尽版本，从而满足不同用户群体的需求。
如先前所讨论，第一答案和第二答案可以对应于客服系统能够在不同的问题输入模态下 提供的答案，不同的问题输入模态则反映出不同的用户需求。本公开实施例中，用户输入问题的问题输入模态可以基于各种方式来确定，以下举例说明确定问题输入模态的几种示例方式。
<示例1-电子设备的设备类型和/或环境参数>
如前所讨论的，用户可以使用其持有的电子设备来输入问题，该电子设备至少可以包括手机、笔记本电脑、平板电脑、智能音箱、个人数字助手等。在本示例中，可以基于用户输入问题时所使用的电子设备的设备类型和/或环境参数，来确定问题输入模态，以便将不同的答案对应于可能出现的问题输入模态。
一方面，电子设备的类型可以指示该电子设备是否支持显示视频/图片/文字等显示功能、是否支持播放语音的功能等，因而能够据此判断该电子设备是否能够支持呈现以某种类型表示的答案，以避免用户得到其电子设备不能够显示/播放的答案，而降低用户满意度。例如，当用户通过智能音箱通过语音方式输入问题时，可以确定该电子设备的类型为智能音箱，不具有显示屏但具有播放声音的能力，在这样的情况下，可以将问题输入模态确定为使用仅具有声音播放功能的电子设备来输入问题的模态，据此可以将声音类型的答案对应于客服系统能够在该问题输入模态下提供的答案。
另一方面，电子设备所处的环境参数可以指示其是否处于无线局域网所覆盖的范围，因而能够据此评估答案传输的带宽限制和流量花费等因素，使得答案能够以经济有效的方式呈现给用户。例如，当用户的电子设备处于未连接到无线局域网的状态时，考虑到手机的流量费用较高，用户可能不期望花费过多流量来获得答案，在这样的情况下，可以将问题输入模态确定为使用未连接无线局域网的电子设备输入问题的模态，据此可以将数据量小的文字类型答案(例如，文字类型的答案)对应于客服系统能够在该问题输入模态下提供的答案。另外，同样是考虑到答案传输的带宽限制和流量花费等因素，可以将同一类型但内容更精简的答案对应于客服系统能够在该问题输入模态下提供的答案。
虽然以上描述了电子设备是否处于无线局域网所覆盖的范围作为其环境参数的示例，本领域技术人员可以理解，还可以考虑电子设备的其他环境参数，包括但不限于：电子设备周围的噪声水平、所处的位置、电子设备运行状态等。比如，当噪声水平低并且指示用户位于图书馆等安静场所时，可以将无声音的答案类型对应于客服系统能够在该问题输入模态下提供的答案；又比如，当电子设备正在运行导航软件而不便于观看电子设备时，可以将声音类型答案对应于客服系统能够在该问题输入模态下提供的答案，等等。
<示例2-用户的知识背景和/或个人偏好>
发明人认识到，不同的用户的知识背景和个人偏好并不相同，因此如何考虑这些因素来 提供合适答案同样是至关重要的。针对这一情况，可以基于用户的知识背景和/或个人偏好，来确定问题输入模态，以便将不同的答案对应于可能出现的问题输入模态。
一方面，可以基于用户通过电子设备输入的识别信息，确定用户的身份，进而确定身份被确定的用户的知识背景。用户的知识背景可以是用户事先通过调查问卷提供的学历信息、工作经验、技术能力等信息，据此可以估计用户对问题答案的理解能力。如果用户是该技术问题领域的技术人员，可以将问题输入模态确定为该领域的技术人员输入问题的模态，在这样的情况下，可以将类型简单的答案(例如，文本类型的答案)和/或内容更精简的答案对应于客服系统能够在该问题输入模态下提供的答案，从而节约用户的时间成本和网络开销；如果用户对该技术问题的领域的知识较少，可以将问题输入模态确定为非相关领域人员输入问题的模态，在这样的情况下，可以将类型生动的答案(例如，视频类型的答案)和/或内容更详尽的答案对应于客服系统能够在该问题输入模态下提供的答案，从而使得用户充分理解该问题的答案内容。
另一方面，可以基于用户通过电子设备输入的识别信息，确定用户的身份，进而确定身份被确定的用户的个人偏好，据此可确定不同的问题输入模态。替代地，用户的个人偏好可以是用户在其电子设备上设定的偏好信息，据此可确定不同的问题输入模态。偏好信息可以是用户事先通过调查问卷提供的更偏好于观看某种类型答案、是否有充足时间学习知识、是否对某领域问题特别关注等信息。例如，用户的个人偏好可以指示该用户喜欢更加生动的语音/视频等问答方式获取答案，据此可以将问题输入模态确定为交互型用户输入问题的模态，在这样的情况下，可以将类型生动的语音/视频等类型的答案对应于客服系统能够在该问题输入模态下提供的答案。
<示例3-问题的答案内容>
申请人认识到，用户输入的各种问题可能对应于不同结构(例如，顺承结构、非顺承结构等)的答案内容，而不同结构的答案内容可能分别适合用不同类型的答案进行表示。在本示例中，可以基于用户输入的问题的答案内容的分析，来确定问题输入模态，以便将不同类型的答案对应于可能出现的问题输入模态。
结合以上标准问答示例的答案内容描述，以上示例1和2中的答案内容中含有较多操作步骤，则表明该答案内容为操作类，据此可以将问题输入模态确定为输入操作相关问题的模态，在这样的情况下，可以将视频类型的答案对应于客服系统能够在该问题输入模态下提供的答案，从而更有助于用户直观地理解和掌握各操作步骤的顺序。又例如，以上示例3-5的答案内容是对定义的解释和/或事实的阐明，则表明该答案内容为定义/事实类内容，据此可以将问题输入模态确定为输入非操作相关问题的模态，在这样的情况下，用文本形式可能就足 以使用户理解，因此可以将文本类型的答案对应于客服系统能够在该问题输入模态下提供的答案。关于如何根据对答案内容分析确定答案内容为操作/定义/事实类的方法，将在下文中予以描述。
以上描述了基于用户用来输入问题的电子设备的设备类型和/或环境参数、用户的知识背景和/或个人偏好、问题的答案内容等因素来确定问题输入模态的示例。本领域技术人员可以理解，根据实际需求，还可以基于其他因素来确定问题输入模态，据此可以将不同的答案对应于可能出现的不同问题输入模态，以供推送给具有不同需求的用户。
返回图1，如结合图1所述描述的，该实施例可以接收用户输入的问题，并确定与用户输入的问题相对应的第一答案，并且将所述第一答案提供给用户，由此可以使用户通过第一答案进行学习以获得知识。可以理解，该实施例的步骤S102中可以通过多种方式确定与用户输入的问题相对应的第一答案，下面，将参照图2A和图2B描述图1的步骤S102中确定与用户输入的问题相对应的第一答案的几种示例性方法。
首先，参考图2A描述确定与用户输入的问题相对应的第一答案的一种示例性方法，该方法包括如下步骤：
步骤S201，获得与用户输入的问题相对应的多个答案。如先前所讨论，第一答案可以是从预先构建的多个答案中选择的，在此情况下，可以首先获得与用户输入的问题相对应的预先构建的多个答案，以供后续筛选出适合推送给用户的第一答案。需说明的是，本实施例中的多个答案可以是预先构建的以不同答案类型表示的多个答案，或者具有不同内容的多个答案。
步骤S203，从所述多个答案中确定所述第一答案。在该步骤中，可以根据各种标准，从获得的多个答案中确定第一答案。作为一种可能的示例，类似于先前关于如何确定问题输入模态所描述的，可以基于用户用来输入问题的电子设备的设备类型和/或环境参数、用户的知识背景和/或个人偏好、问题的答案内容等因素，从已获得的多个答案中确定第一答案。
接下来，参考图2B描述确定与用户输入的问题相对应的第一答案的另一种示例性方法，该方法包括如下步骤：
步骤S202，确定用来表示答案的答案类型。可以理解，上述步骤S101中根据用户输入的问题所确定的第一答案是有待将答案内容以某种类型呈现给用户的，而选择适当类型的答案可以满足用户的需求。类似于步骤S203中从已获得的多个答案中确定第一答案所描述的，可以基于用户用来输入问题的电子设备的设备类型和/或环境参数、用户的知识背景和/或个人偏好、问题的答案内容等因素，确定用来表示答案的答案类型，以供后续从后台服务器获得该类型的答案。
步骤S204，获得以所确定的答案类型表示的答案作为所述第一答案。例如，可以从后台数据库预先构建的多个答案中选择类型为所确定的答案类型的第一答案，以将其提供给用户；又例如，当后台数据库预先构建的多个答案中不含有所确定的答案类型的答案时，可以实时构建以所确定的答案类型表示的答案，并将其提供给用户，本公开的实施例不以此为限。
以上结合图2A和图2B描述了确定与用户输入的问题相对应的第一答案的几种示例性方法，并且在图2B的步骤S202中描述了可以基于各种因素确定用来表示答案的答案类型以供后续获得该类型的答案推送给用户。下面，将参照图3具体描述图2B的步骤S202中基于问题的答案内容确定用来表示答案的答案类型的一种示例性方法。如图3所示，该实施例的方法可以包括如下步骤：
步骤S301，对所述答案的答案内容进行特征提取。可以理解，所述答案内容是包含能够有效地解答用户疑问的标准内容，并且是有待以某种类型进行表示以供推送给用户的。在该步骤中，所提取的特征包括顺承符号特征、顺承词特征、交互操作词特征和交互操作序列长度特征中的至少一个。
以下结合上述标准问答示例1-5中的答案内容，介绍顺承符号特征、顺承词特征、交互操作词特征和交互操作序列长度特征。顺承符号特征可以是如标准问答示例1中的顺承符号“->”出现的次数，以表征各个操作步骤具有的前后顺承关系。顺承词特征可以是如顺承词频，以表征诸如标准问答示例2中的“首先”、“然后”等顺承词出现的频率。交互操作词特征可以是有关系统软件和应用软件中操作类词频，以表征诸如标准问答示例1和2中的“设置”、“VPN”、“更多”等交互操作类词出现的频率。交互操作序列长度特征可以是系统软件和应用软件中操作类词连续出现的次数，以表征交互操作序列的总长度和复杂度。可以理解的是，以上所列举的特征仅为示例，还可以根据实际需求采用不同的特征提取方法提取出能够反映答案内容的不同种类特征。
步骤S302，根据所提取的特征，确定所述答案内容的结构。例如，根据所提取的特征，可以将所述答案内容的结构确定为顺承结构或者非顺承结构。例如，可以将标准问答示例1和2的答案内容确定为具有顺承结构，将问答示例3-5中的答案内容确定为具有非顺承结构。虽然该步骤中描述了将答案确定为顺承结构或者非顺承结构这两种结构类别，本领域技术人员可以理解，在步骤S301中采用不同的特征提取方法的情况下，可以在步骤S302中将答案划分为更多的不同结构类别。
步骤S303，基于所述答案内容的结构，确定分别以多个不同答案类型来表示答案的各个优先级得分。例如，考虑到具有顺承结构的答案内容包含较多操作步骤，以视频类型表示则更有助于用户理解，因此，针对顺承结构的答案内容，可以给予视频类型较高的优先级得分， 而给予其他类型较低的优先级得分。又例如，考虑到非顺承结构的答案内容大多是解释性或定义性的说明，以文字形式表示就足以使用户理解并且能节省用户的浏览时间和传输带宽，因此，针对非顺承结构的答案内容，可以给予文本类型较高的优先级得分，而给予其他类型较低的优先级得分。
作为非限制性的示例，该优先级得分可以是优先级得分数值(例如，0～100区间的得分值)，还可以是优先级的排序(例如，第一优先级、第二优先级等)。例如，可以将顺承结构的答案内容的视频类型设定为第一优先级，而将其余类型设定为第二优先级；将非顺承结构的答案内容的文本类型设定为第一优先级，而将其余类型设定为第二优先级。
步骤S304，基于所述各个优先级得分，确定用来表示答案的答案类型。在该步骤中，在获得了各个优先级得分之后，通过对各个得分加以比较可以确定出适合推送给用户的答案类型，以便最终以该类型来表示答案内容并且推送给用户，使得答案更容易被用户所理解和接受。例如，可以以优先级得分最高的类型来表示答案，以呈现给用户进行学习。
根据本公开实施例的答案提供方法，可以通过对答案的内容进行分析确定以不同答案类型表示答案的各个优先级得分，并且据此自动确定最适合推送给用户的答案类型，由此避免大量人工标注所导致的资源浪费和效率低的问题，并且使得用户更好地理解答案内容，提高用户对答案的满意度。
以上结合图3描述了基于答案的内容确定以不同答案类型表示答案的各个优先级得分，并且在确定了各个优先级得分之后，预期以最高优先级得分的答案类型来表示答案的可能实现方式。然而，实际中可能存在以下情况：对于用户所输入的问题，事先已经构建了几种类型的答案，但是目前尚未以优先级得分最高的类型构建答案；或者，用户输入的问题是该领域中新出现的问题，目前尚未构建出任何类型的问题答案。在这样的情况下，由于尚未以优先级得分最高的类型构建答案，因而暂时无法以该类型表示答案并向用户推送。下面，将参照图4描述图3的步骤S304中基于所述各个优先级得分确定用来表示答案的答案类型的一种示例性方法。如图4所示，该实施例的确定各个优先级得分的方法可以包括如下步骤：
步骤S401，确认是否预先构建了以优先级得分最高的答案类型表示的答案。
步骤S402，当预先构建了以优先级得分最高的答案类型表示的答案时，确认以所述优先级得分最高的答案类型来表示答案。在此情况下，优先级得分最高的答案类型已经预先构建，则客服系统可以获得该类型的答案作为第一答案，以提供给用户进行学习。
步骤S403，当尚未构建以优先级得分最高的答案类型表示的答案时，决定是否要构建以优先级得分最高的答案类型表示的答案。在此情况下，优先级得分最高的答案类型尚未构建，因此客服系统可以考虑各种因素来决定是否要去构建该优先级得分最高的答案类型的答案。 如果决定了要构建以优先级得分最高的答案类型表示的答案，那么，客服系统便可以将构建好的答案作为第一答案提供给用户。如果决定不构建以优先级得分最高的答案类型表示的答案，那么可以进一步决定是否构建以优先级得分第二的答案类型表示的答案、决定是否构建以优先级得分第三的答案类型表示的答案，以此类推，从而最终确定用来表示答案的类型，之后可以将构建好的答案作为第一答案提供给用户。
一般地，以视频类型来构建答案的成本相对较高，以文本类型来构建答案的成本相对较低，以语音和图片类型来构建答案的成本介于二者之间。考虑到客服系统的运营成本和用户满意度，高频问题可能更值得花费资源构建较高成本类型的答案，而低频问题花费较高成本去构建答案则会造成资源的浪费。因此，在构建答案时如何在构建成本和答案使用率二者之间权衡至关重要。如先前所讨论的，可以考虑各种因素来决定是否要去构建该优先级得分最高的答案类型的答案，下面，将参照图5描述图4的步骤S403中决定是否要构建以优先级得分最高的答案类型表示的答案的一种示例性实现方法。如图5所示，该方法可以包括如下步骤：
步骤S501，计算以优先级得分最高的答案类型构建所述答案的构建成本。对于具有不同结构答案内容的答案，其构建成本的计算方法有所不同。以下将分别顺承结构的答案和非顺承结构的答案作为示例，描述答案的构建成本。
<顺承结构答案的构建成本>
当答案内容的结构被确定为顺承结构时，计算以优先级得分最高的答案类型来构建答案的构建成本可以包括如下步骤：
(A)将所述答案内容划分为多个操作步骤，计算以优先级得分最高的答案类型构建所述各个操作步骤的各个步骤构建成本。
(B)基于各个步骤构建成本，计算所述构建成本。
如上所述，顺承结构答案是包含具有顺承关系的多个操作步骤的答案，可以将顺承结构的答案内容表示为A＝{s 1,s 2,…s n}，其中s i表示答案内容中的第i个步骤，并且可以计算以优先级得分最高的答案类型来构建各个操作步骤s 1～s 4的各个步骤构建成本。进一步地，基于计算出的各个步骤s 1～s 4的各个步骤构建成本，可以得到以优先级得分最高的答案类型构建整个顺承结构答案的构建成本。当然，以上考虑的是对于每个各个操作步骤都去计算其步骤构建成本的情况，实际上，如果其中某些操作步骤的素材预先制作过而无需重复构建时，可以将那些操作步骤的步骤构建成本设定为零或者设定相对较低值，以准确地反映出实际场景中的答案制作费用。
例如，上述标准问答示例1的答案内容中包括4个步骤，分别为s 1＝进入设置、s 2＝关于 手机、s 3＝状态、s 4＝IMEI信息。可以分别计算出以优先级得分最高的答案类型(例如，视频类型)构建以上4个操作步骤的步骤构建成本，进而计算以该类型来构建整个答案的构建成本。
作为计算各个步骤构建成本的一种示例性实现方式，可以在预先构建的步骤成本知识库中检索，以查找出各个操作步骤所对应的各个步骤构建成本。在该示例中，预先构建的步骤成本知识库可以是基于操作步骤的难易程度来构建的。例如，该预先构建的步骤成本知识库涉及手机领域中可分解出的大多数操作步骤，诸如“进入设置”、“打开相机”、“关于手机”等常规操作步骤，该步骤成本知识库可以以单个操作步骤为单位，管理以各种不同类型(诸如，视频、文本等)构建该单个步骤的步骤构建成本。因此，可以根据所要检索的各个操作步骤的关键词从该步骤成本知识库中进行检索，以取得对应于各个操作步骤的各个步骤构建成本，以进行后续的整个答案的构建成本计算。
以下将结合图6A、图6B和图6C描述构建步骤成本知识库的一种示例性实现方法。如图6A所示，该方法可以包括如下步骤：
步骤S601，预先确定以至少一种答案类型构建至少一个顺承结构的答案内容的各个操作步骤的步骤构建成本。以上所述的至少一种答案类型至少涵盖待计算的优先级得分最高的答案类型，并且以上所述的至少一个顺承结构的答案包括诸如手机领域中常见操作交互类问题所对应的多个顺承结构答案。在对以上所述的至少一个顺承结构的答案进行步骤原子化之后，则可以获得各个原子化的操作步骤，以供步骤成本知识库以步骤为单位对构建成本进行管理。取决于不同应用场景，步骤成本知识库可以涵盖不同种类的答案类型，并且其构建过程中可以收集本领域中不同数量的顺承结构的答案，从而构建出不同知识储备等级的步骤成本知识库。
应理解的是，不同答案类型对应着不同等级的构建成本，例如构建视频类型的答案的成本一般要高于构建文本类型的答案。因此，对于不同的答案类型可以赋予不同的基准成本，例如，可以赋予视频类型最高的基准成本，而对于文本类型赋予最低的基准成本。在确定了基准成本之后，可以基于各个操作步骤的难易程度并且结合上述基准成本，最终确定各个步骤构建成本。
以视频类型为例，对于可通过系统软件和应用软件完成的操作步骤，例如可以简单地通过对软件操作截屏或录屏等方式获得构建视频答案的素材的情况，可以在基准成本的基础上结合相对较低的成本。对于需要系统软件和应用软件以外的操作步骤，例如对于需要连接笔记本电脑进行操作因而获得制作视频的素材相对困难的情况，可以在基准成本的基础上结合相对较高的成本。当然，以上描述的计算成本的方式仅为示例性说明，本领域技术人员可以 根据实际情况和制作人员的制作能力，对于不同的操作步骤赋予不同的操作步骤构建成本。
步骤S602，基于所述至少一个顺承结构的答案的各个操作步骤，构建操作步骤集合。在该步骤中，可以将步骤S601中经原子化的各个操作步骤进行组合并且删除重复项，从而得到操作步骤的集合SS＝{ss 1,ss 2,ss 3,…ss N}。该操作步骤的集合涵盖了手机领域中常见的操作步骤，进而，可以从集合SS中查找到与如上述标准问答示例1中各操作步骤s 1～s 4相对应的各个操作步骤。如上所述，当所收集的顺承结构的答案的数量越多时，操作步骤集合所涵盖的操作步骤跨越范围越大，该步骤成本知识库的知识储备等级越高。
步骤S603，以所述操作步骤集合中的单个操作步骤为单位，构建步骤成本知识库。例如，所构建的步骤成本知识库如图6B所示，其中操作步骤的集合SS中的每个操作步骤ss i均具有与之对应的、以一种或多种类型构建答案的成本Css i。
此外，本公开的发明人认识到，在该步骤的集合中，各个操作步骤的重要度并不相同，而步骤重要度同样是衡量该步骤的优先级的重要因素。例如，如果某一操作步骤的重要度较高，则说明该操作步骤在手机领域操作中是较为常用且必要的，因此构建出该操作步骤的素材则在后续其他答案的构建过程中多次使用，以避免重复制作个别操作步骤所导致的资源浪费。换句话说，某一操作步骤的重要度较高，则表明更值得去构建该操作步骤的答案，在此情况下，可以考虑人为地在其步骤构建成本上加权，即便其实际构建成本可能稍高，但考虑到构建出来的素材可能会被多次使用，仍对其加权而将其视为低构建成本的。因此，基于步骤重要度对已计算出的各个步骤构建成本进行加权，则可以更准确地反映出实际制作场景中的各个步骤的最终成本。以下结合图6C描述对步骤成本知识库中的各个操作步骤的步骤构建成本加权的方法的示例。
在如上所述步骤S603中构建的成本知识库中，包含多个经原子化的操作步骤ss 1～ss N。各个原子化步骤ss 1～ss N在所收集的原始顺承结构答案中与其他步骤具有前序/后序两种顺承关系，基于各个原子化步骤ss 1～ss N在原始顺承结构答案中所具有的前序/后序关系，可以构建以所有原子步骤为节点、以前序/后序关系为边的步骤有向图。进一步地，基于步骤有向图可以计算出单个原子化步骤的重要度。例如，作为计算各步骤的重要度的一种示例性方法，可以采用PageRank算法来实现。
如图6C所示，以步骤ss 1～ss 3为例，各个步骤ss 1～ss 3均与其他步骤之间具有前序/后序关系，例如ss 3为ss 2的后续，ss 2为ss 1的后续。虽然图6C示意性地示出了每个步骤与另外的一个步骤之间具有前序/后序关系，但是本领域技术人员可以理解的是，由于每个步骤均可能来源于所收集到的不同的原始顺承结构答案，因而每个步骤可以与两个甚至更多个步骤之间具有前序/后序关系，本发明不以此为限。进而，可以对步骤成本知识库中的各个操作步骤的步 骤构建成本加权，并以加权后的步骤构建成本更新所述步骤成本知识库。例如，假设图6C中步骤ss i的重要度计算为SRss i，则步骤ss i的加权成本可以为Css i/SRss i。
<非顺承结构答案的构建成本>
顺承结构答案是指不包含多个具有顺承关系的操作步骤的答案，其示例可以包括：不具有顺承结构的操作类答案、事实类答案和定义类答案等。对于不同类型的非顺承结构答案，可以采用不同的方法计算其答案的构建成本。以下描述计算非顺承结构答案的构建成本的示例性方法。
当答案内容的结构被确定为非顺承结构时，计算以优先级得分最高的答案类型来构建答案的构建成本可以包括如下步骤：
(A)确定所述答案内容是否为操作类内容。例如，可以采用特征提取方法提取答案中的交互操作词特征等，并且基于所提取的特征结合语义分析来确定该答案是否为操作类答案。以标准问答示例3为例，可以提取其中的“下载”等交互操作词特征，并且结合其上下文语义分析确定该答案内容为操作类答案。
(B)当步骤(A)中确定所述答案内容是操作类内容时，基于所述操作类内容对应的操作难易程度计算所述构建成本。例如，与以上结合顺承结构答案成本构建所描述的类似地，可以确定以优先级得分最高的答案类型构建该答案的基准成本，并且进一步结合该答案中的步骤的操作难易程度，确定以优先级得分最高的答案类型构建该答案的构建成本。
(C)当步骤(A)中确定所述答案内容不是操作类内容时，进一步确定所述答案内容是否为事实类内容或定义类内容。例如，可以采用特征提取方法提取答案中的特征等，并且基于所提取的特征确定该答案是否为事实类内容或定义类内容。例如，通过特征提取和识别方法，可以确定示例4中的答案为定义类内容，示例5中的答案为事实类内容。
(D)当步骤(C)中确定所述答案内容为事实类内容或定义类内容时，基于所述答案内容的复杂性和相关用户技术背景计算所述构建成本。例如，有些技术术语或定义比较复杂或者对于广大用户群体比较陌生，需要配合大量的讲解才能使得用户掌握其含义，因此需要准备较多的制作素材，可以对其设定较高的构建成本，例如以上标准问答示例4。而有些答案比较简单，则可以对其设定较低的构建成本，例如以上标准问答示例5。
返回到图5，上述决定是否要构建以优先级得分最高的答案类型表示的答案的示例方法还包括：步骤S502，预测所述答案的使用率。以下结合图7对预测答案的使用率的示例性方法进行描述，预测答案的使用率的示例性方法可以包括如下步骤：
(A)对至少一个用户的多个历史问题进行分类。例如，可以根据用户问题的意图进行粗略分类，将用户的历史问题分为以下各个类别：“系统更新”、“电池”、“屏幕显示”、“数据传 输”、“拍照功能”等。对用户的多个历史问题的分类方法可以使用传统的分类器如SVM和神经网络分类模型CNN等，本公开的实施例不以此为限。如图7所示，可以将对多个用户的历史问题进行分类，例如将问题q1、q2和q3分类为系统更新类别，将问题q4、q5分类为电池类别。
(B)确定多个历史问题中属于标准问题的分类的至少一个历史问题与所述标准问题是否匹配，其中所述标准问题与所述答案相对应。如图7所示，可以分别确定问题q1、q2和q3与属于同一类别“系统更新”的标准问题Q1和Q2是否匹配，且每个标准问题Q1和Q2都分别具有与之对应的唯一答案A1和A2，而每个答案A1和A2即为有待计算其使用率的答案。类似地，可以确定问题q4和q5与标准问题Q3是否匹配。例如，作为判断历史问题和标准问题是否匹配的示例性方法，可以使用预先训练好的词向量作为初始值对用户的历史问题和标准问题进行向量化表示，然后使用CNN、LSTM等算法对用户的历史问题和标准问题进行编码得到句向量，根据句向量确定用户的历史问题和标准问题是否相匹配。图7示出了用户的历史问题和标准问题的匹配结果，其中历史问题q1与标准问题Q2相匹配，历史问题q2与标准问题Q1相匹配，历史问题q3与标准问题Q2相匹配，历史问题q4与标准问题Q3相匹配，历史问题q5与标准问题Q3相匹配。
(C)基于多个历史问题中与所述标准问题相匹配的历史问题的数量，预测所述答案的使用率。如前所述，标准问题分别与待计算其使用率的答案一一对应，根据相匹配的历史问题的数量，便可以确定该标准问题被提问的次数，据此反映出该答案的使用率。例如，结合图7的用户的历史问题和标准问题的匹配结果，确定与标准问题Q1相匹配的历史问题数量为1，与标准问题Q2相匹配的历史问题数量为2，与标准问题Q3相匹配的历史问题数量为2，进而可以依据用户的历史问题、标准问题和有待计算使用率的答案之间的对应关系，估计标准答案A1～A3被使用的频率。例如，可以将对应到该答案A1的用户历史问题数占历史问题总数的比例作为答案A1的使用率。
以上描述了基于历史问题的数量来预测所述答案的使用率的实施例。可以理解，还可以采用其他方式进行预测。例如，当某一问题及其答案是该领域中比较新鲜的问题时，与该答案对应的历史问题的数量可能相对有限，在此情况下，则可以基于对该问题的热门趋势、感兴趣群体数量等因素的估计，来预测所述答案的使用率的实施例。
返回到图5，该方法还包括：步骤S503，基于所述构建成本和使用率，决定是否构建以优先级得分最高的答案类型表示的答案。假设对于上述答案A的构建成本为C(A)且使用率为U(A)，可以计算二者之间的比值U(A)/C(A)。进一步地，将该比值与预定的阈值相比较，当该比值大于预定的阈值时，说明值得以优先级得分最高的答案类型来表示答案，则可以决定 构建以优先级得分最高的答案类型表示的答案；当该比值小于预定的阈值时，说明不值得以优先级得分最高的答案类型来表示答案，则可以决定不构建以优先级得分最高的答案类型表示的答案。该预定的阈值可以根据用户需求进行设定，本发明不以此为限。
另外，考虑到客服系统的运营成本有限，可能不会针对每个答案都去构建以优先级得分最高的答案类型表示的答案。在此情况下，可以计算有待构建的多个答案的各个答案类型的各个构建成本，并且基于制作经费等因素从中筛选出实际要去构建的多个答案，并且对于未被筛选的答案，可以考虑以较低构建成本的类型来构建，来保证客服系统的正常运营。
除此之外，用户对答案内容的总体满意度评价也是衡量答案的重要因素。鉴于此，可以根据用户平均满意度得分对上述比值进行加权，并且以加权后的比值来决定是否构建以优先级得分最高的答案类型表示的答案。例如，假设用户对答案内容的总体满意度为SF(A)，可以对上述比值比值U(A)/C(A)乘以加权系数SF(A)，并据此决定是否构建以优先级得分最高的答案类型表示的答案。
以上结合附图描述了根据本公开实施例的答案提供方法，可以基于用户输入的问题确定并且提供相对应的答案，并且还能够在不同的问题输入模态下给用户提供不同的答案，由此满足用户对答案的不同需求，提高用户对答案的满意度。另外，通过对答案的内容进行分析确定以不同答案类型表示答案的各个优先级得分，并且据此自动确定适合推送给用户的答案类型，可以避免大量人工标注，并且使得用户更好地理解答案的内容，提高用户对答案的满意度。除此之外，在决定是否要以某一类型构建答案时，可以考虑该类型答案的构建成本及其使用率，由此兼顾客服系统的运营成本和用户满意度。
下面，将参照图8来描述根据本公开实施例的答案提供设备的框图。图8示出了根据本公开实施例的答案提供设备800的示例性结构框图。
答案提供设备800例如可以是笔记本电脑、服务器、云端数据装置等任何类型的电子装置。该答案提供设备800可以接收用户输入的问题，并且向用户提供相对应的答案。以下仅对答案提供设备800的各单元的主要功能进行描述，而省略以上已经描述过的细节内容。如图8所示，该答案提供设备800可以包括问题接收单元801、答案确定单元802、答案提供单元803。此外，根据需要，答案提供设备800还可以包括图8中未示出的其他组件，比如显示器、输入/输出设备等。
问题接收单元801可以接收用户输入的问题。例如，用户可以使用其持有的电子设备或者该答案提供设备本身通过各种方式向客服系统输入想要询问的问题。用户输入问题的方式可以包括但不限于：打字输入、语音输入、在问题列表中选择问题等。
答案确定单元802可以确定与用户输入的问题相对应的第一答案。例如，在接收到用户 输入的问题后，客服系统可以从后台数据库获得与所输入的问题相对应的第一答案。需说明的是，本实施例中的第一答案可以从是后台数据库预先构建的多个答案中选择的，或者是基于用户所输入的问题而实时构建的，本公开的实施例不以此为限。
答案提供单元803可以将所述第一答案提供给用户。例如，在确定了第一答案之后，客服系统便可将其推送给用户用来输入问题的电子设备。需说明的是，虽然描述了答案提供单元803将第一答案提供给用户，但本实施例还考虑了不同的问题输入模态下的不同用户需求，因此本实施例中用户输入的问题在不同的问题输入模态下可以对应于与第一答案不同的答案，由此答案提供单元803便能够在可能出现不同的问题输入模态时向用户推送合适的答案。在本实施例中，所述问题至少还对应于第二答案，所述第二答案是在不同的问题输入模态下提供给用户的答案。例如，用户输入的问题可能是客服系统中经常遇到或者预计将要遇到的问题，考虑不同的问题输入模态下不同的用户需求，本实施例可以将该问题对应于多个不同答案，以便答案提供单元803能够在可能出现的不同的问题输入模态下选择合适的答案来向用户推送，使得执行该答案提供方法的客服系统具有在可能出现的不同的问题输入模态下提供合适的答案的能力。
需说明的是，答案提供设备800还可以执行如先前结合图1-图7描述的答案提供方法中的任何处理，具体细节在此不予赘述。
根据本公开的实施例的模块、子模块、单元、子单元中的任意多个、或其中任意多个的至少部分功能可以在一个模块中实现。根据本公开实施例的模块、子模块、单元、子单元中的任意一个或多个可以被拆分成多个模块来实现。根据本公开实施例的模块、子模块、单元、子单元中的任意一个或多个可以至少被部分地实现为硬件电路，例如现场可编程门阵列(FPGA)、可编程逻辑阵列(PLA)、片上系统、基板上的系统、封装上的系统、专用集成电路(ASIC)，或可以通过对电路进行集成或封装的任何其他的合理方式的硬件或固件来实现，或以软件、硬件以及固件三种实现方式中任意一种或以其中任意几种的适当组合来实现。或者，根据本公开实施例的模块、子模块、单元、子单元中的一个或多个可以至少被部分地实现为计算机程序模块，当该计算机程序模块被运行时，可以执行相应的功能。
例如，问题接收单元801、答案确定单元802、答案提供单元803中的任意多个可以合并在一个模块中实现，或者其中的任意一个模块可以被拆分成多个模块。或者，这些模块中的一个或多个模块的至少部分功能可以与其他模块的至少部分功能相结合，并在一个模块中实现。根据本公开的实施例，问题接收单元801、答案确定单元802、答案提供单元803中的至少一个可以至少被部分地实现为硬件电路，例如现场可编程门阵列(FPGA)、可编程逻辑阵列(PLA)、片上系统、基板上的系统、封装上的系统、专用集成电路(ASIC)，或可以通过 对电路进行集成或封装的任何其他的合理方式等硬件或固件来实现，或以软件、硬件以及固件三种实现方式中任意一种或以其中任意几种的适当组合来实现。或者，问题接收单元801、答案确定单元802、答案提供单元803中的至少一个可以至少被部分地实现为计算机程序模块，当该计算机程序模块被运行时，可以执行相应的功能。
以上结合附图描述了根据本公开实施例的答案提供设备，根据本公开实施例的答案提供设备，可以基于用户输入的问题确定并且提供相对应的答案，并且还能够在不同的问题输入模态下给用户提供不同的答案，由此满足用户对答案的不同需求，提高用户对答案的满意度。另外，通过对答案的内容进行分析确定以不同答案类型表示答案的各个优先级得分，并且据此自动确定适合推送给用户的答案类型，可以避免大量人工标注，并且使得用户更好地理解答案的内容，提高用户对答案的满意度。除此之外，在决定是否要以某一类型构建答案时，可以考虑该类型答案的构建成本及其使用率，由此兼顾客服系统的运营成本和用户满意度。
本公开的另一实施例还提供了一种计算机可读存储介质，用于存储非暂时性计算机可读指令，当所述非暂时性计算机可读指令由计算机执行时可以执行本公开的实施例的答案提供方法。该计算机可读存储介质可以是上述实施例中描述的设备/装置/系统中所包含的；也可以是单独存在，而未装配入该设备/装置/系统中。前述的存储介质包括但不限于：便携式计算机磁盘、硬盘、随机访问存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本公开中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。
附图中的流程图和框图，图示了按照本公开各种实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段、或代码的一部分，上述模块、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意，在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个接连地表示的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图或流程图中的每个方框、以及框图或流程图中的方框的组合，可以用执行规定的功能或操作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
本领域技术人员可以理解，本公开的各个实施例和/或权利要求中记载的特征可以进行多种组合和/或结合，即使这样的组合或结合没有明确记载于本公开中。特别地，在不脱离本公开精神和教导的情况下，本公开的各个实施例和/或权利要求中记载的特征可以进行多种组合和/或结合。所有这些组合和/或结合均落入本公开的范围。
尽管已经参照本公开的特定示例性实施例示出并描述了本公开，但是本领域技术人员应该理解，在不背离所附权利要求及其等同物限定的本公开的精神和范围的情况下，可以对本公开进行形式和细节上的多种改变。因此，本公开的范围不应该限于上述实施例，而是应该不仅由所附权利要求来进行确定，还由所附权利要求的等同物来进行限定。

Claims (10)

  1. 一种答案提供方法，包括：
    接收用户输入的问题；
    确定与用户输入的问题相对应的第一答案；以及
    将所述第一答案提供给用户，
    其中，所述问题至少还对应于第二答案，所述第二答案是在不同的问题输入模态下提供给用户的答案。
  2. 根据权利要求1所述的方法，其中，所述第一答案和所述第二答案是以不同答案类型表示的答案或者具有不同内容的答案，所述问题输入模态基于以下中的一个或多个来确定：
    用户用来输入问题的电子设备的设备类型和/或环境参数；
    用户的知识背景和/或个人偏好；以及
    所述问题的答案内容。
  3. 根据权利要求1所述的方法，其中，所述确定与用户输入的问题相对应的第一答案包括：
    获得与用户输入的问题相对应的多个答案，所述多个答案是以不同答案类型表示的多个答案或者具有不同内容的多个答案；以及
    从所述多个答案中确定所述第一答案。
  4. 根据权利要求1所述的方法，其中，所述确定与用户输入的问题相对应的第一答案包括：
    确定用来表示答案的答案类型；以及
    获得以所确定的答案类型表示的答案作为所述第一答案。
  5. 根据权利要求4所述的方法，其中，所述确定用来表示答案的答案类型包括：
    对所述问题的答案内容进行特征提取，其中所提取的特征包括顺承符号特征、顺承词特征、交互操作词特征和交互操作序列长度特征中的至少一个；
    根据所提取的特征，确定所述答案内容的结构，其中所述答案内容的结构至少包括顺承结构和非顺承结构；
    基于所述答案内容的结构，确定分别以多个不同答案类型来表示答案的各个优先级得分；以及
    基于所述各个优先级得分，确定用来表示答案的答案类型。
  6. 根据权利要求5所述的方法，其中，所述基于所述各个优先级得分确定用来表示答案 的答案类型包括：
    确定是否预先构建了以优先级得分最高的答案类型表示的答案，其中
    当预先构建了以优先级得分最高的答案类型表示的答案时，确认以所述优先级得分最高的答案类型来表示答案，
    当尚未构建以优先级得分最高的答案类型表示的答案时，决定是否要构建以优先级得分最高的答案类型表示的答案。
  7. 根据权利要求6所述的方法，其中，所述决定是否要构建以优先级得分最高的答案类型表示的答案包括：
    计算以优先级得分最高的答案类型来构建答案的构建成本；
    预测所述答案内容的使用率；以及
    基于所述构建成本和使用率，决定是否构建以优先级得分最高的答案类型表示的答案。
  8. 根据权利要求7所述的方法，其中，
    当所述答案内容的结构为顺承结构时，所述计算以优先级得分最高的答案类型来构建答案的构建成本包括：
    将所述答案内容划分为多个操作步骤，计算以优先级得分最高的答案类型构建各个操作步骤的各个步骤构建成本；以及
    基于各个步骤构建成本，计算所述构建成本，
    其中，所述各个步骤构建成本是从一预先构建的步骤成本知识库中查找而确定的，所述预先构建的步骤成本知识库基于操作步骤的难易程度来构建，
    当所述答案内容的结构为非顺承结构时，所述计算以优先级得分最高的答案类型来构建答案的构建成本包括：
    确定所述答案内容是否为操作类内容，其中，
    当确定所述答案内容是操作类内容时，基于所述操作类内容对应的操作难易程度计算所述构建成本，
    当确定所述答案内容不是操作类内容时，进一步确定所述答案内容是否为事实类内容或定义类内容，
    当确定所述答案内容为事实类内容或定义类内容时，基于所述答案内容的复杂性和相关用户技术背景计算所述构建成本。
  9. 根据权利要求8所述的方法，其中，所述预测所述答案的使用率包括：
    对至少一个用户的多个历史问题进行分类；
    确定多个历史问题中属于标准问题的分类的至少一个历史问题与所述标准问题是否匹配， 其中所述标准问题与所述答案相对应；以及
    基于多个历史问题中与所述标准问题相匹配的历史问题的数量，预测所述答案的使用率。
  10. 一种答案提供设备，包括：
    问题接收单元，配置为接收用户输入的问题；
    答案确定单元，配置为确定与用户输入的问题相对应的第一答案；以及
    答案提供单元，配置为将所述第一答案提供给用户，
    其中，所述问题至少还对应于第二答案，所述第二答案是在不同的问题输入模态下提供给用户的答案。
PCT/CN2018/115430 2018-06-28 2018-11-14 一种答案提供方法及设备 WO2020000867A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810691095.4 2018-06-28
CN201810691095.4A CN108875055B (zh) 2018-06-28 2018-06-28 一种答案提供方法及设备

Publications (1)

Publication Number Publication Date
WO2020000867A1 true WO2020000867A1 (zh) 2020-01-02

Family

ID=64296580

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/115430 WO2020000867A1 (zh) 2018-06-28 2018-11-14 一种答案提供方法及设备

Country Status (2)

Country Link
CN (1) CN108875055B (zh)
WO (1) WO2020000867A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111460816A (zh) * 2020-03-30 2020-07-28 招商局金融科技有限公司 基于语义特征的数据检索方法、电子装置及存储介质
CN112200624A (zh) * 2020-09-28 2021-01-08 深圳市其乐游戏科技有限公司 商品推荐方法、装置、终端设备以及存储介质
CN112988998A (zh) * 2021-03-15 2021-06-18 中国联合网络通信集团有限公司 应答方法及设备
US20210279605A1 (en) * 2020-03-06 2021-09-09 International Business Machines Corporation Efficiently generating accurate responses to a multi-facet question by a question answering system

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110032628A (zh) * 2019-02-21 2019-07-19 北京奥鹏远程教育中心有限公司 一种用户在线咨询系统和方法
CN110070810A (zh) * 2019-05-27 2019-07-30 珠海幸福家网络科技股份有限公司 一种楼盘讲解方法及楼盘讲解系统
CN110381381B (zh) * 2019-06-21 2021-12-07 深圳市轱辘车联数据技术有限公司 一种视频定制方法、视频定制装置及电子设备
CN110442698B (zh) * 2019-08-06 2021-12-28 思必驰科技股份有限公司 对话内容生成方法及系统
CN111353290B (zh) * 2020-02-28 2023-07-14 支付宝(杭州)信息技术有限公司 一种自动响应用户询问的方法和系统
CN112883177A (zh) * 2021-02-18 2021-06-01 联想(北京)有限公司 一种智能回复方法及装置
CN116821311B (zh) * 2023-08-30 2023-10-31 环球数科集团有限公司 一种基于关联数据和用户行为分析的智能客服系统

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105760417A (zh) * 2015-01-02 2016-07-13 国际商业机器公司 基于个性化用户模型和情境的认知交互式搜索的方法和系统
US20170154626A1 (en) * 2015-11-27 2017-06-01 Samsung Electronics Co., Ltd. Question and answer processing method and electronic device for supporting the same
CN106980626A (zh) * 2016-01-18 2017-07-25 佳能株式会社 处理装置及方法以及问题与答案系统

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104216913B (zh) * 2013-06-04 2019-01-04 Sap欧洲公司 问题回答方法、系统和计算机可读介质
WO2018000270A1 (zh) * 2016-06-29 2018-01-04 深圳狗尾草智能科技有限公司 一种基于用户画像的个性化回答生成方法及系统
CN106227779A (zh) * 2016-07-18 2016-12-14 深圳追科技有限公司 一种客服系统的人机交互方法
CN106572001B (zh) * 2016-10-31 2019-10-11 厦门快商通科技股份有限公司 一种智能客服的对话方法及系统

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105760417A (zh) * 2015-01-02 2016-07-13 国际商业机器公司 基于个性化用户模型和情境的认知交互式搜索的方法和系统
US20170154626A1 (en) * 2015-11-27 2017-06-01 Samsung Electronics Co., Ltd. Question and answer processing method and electronic device for supporting the same
CN106980626A (zh) * 2016-01-18 2017-07-25 佳能株式会社 处理装置及方法以及问题与答案系统

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210279605A1 (en) * 2020-03-06 2021-09-09 International Business Machines Corporation Efficiently generating accurate responses to a multi-facet question by a question answering system
CN111460816A (zh) * 2020-03-30 2020-07-28 招商局金融科技有限公司 基于语义特征的数据检索方法、电子装置及存储介质
CN111460816B (zh) * 2020-03-30 2023-03-31 招商局金融科技有限公司 基于语义特征的数据检索方法、电子装置及存储介质
CN112200624A (zh) * 2020-09-28 2021-01-08 深圳市其乐游戏科技有限公司 商品推荐方法、装置、终端设备以及存储介质
CN112988998A (zh) * 2021-03-15 2021-06-18 中国联合网络通信集团有限公司 应答方法及设备
CN112988998B (zh) * 2021-03-15 2023-06-16 中国联合网络通信集团有限公司 应答方法及设备

Also Published As

Publication number Publication date
CN108875055A (zh) 2018-11-23
CN108875055B (zh) 2021-12-24

Similar Documents

Publication Publication Date Title
WO2020000867A1 (zh) 一种答案提供方法及设备
WO2021036876A1 (zh) 直播辅助数据的提供方法、装置、设备及可读介质
US20190116265A1 (en) Insight based routing for help desk service
TW202018533A (zh) 資料處理模型構建方法、裝置、伺服器和用戶端
US20200285353A1 (en) Apparatus for vision and language-assisted smartphone task automation and method thereof
CN107463700B (zh) 用于获取信息的方法、装置及设备
EP3690644A1 (en) Electronic device and operation method therefor
WO2019062413A1 (zh) 应用程序管控方法、装置、存储介质及电子设备
US10783885B2 (en) Image display device, method for driving the same, and computer readable recording medium
WO2019172704A1 (en) Method for intent-based interactive response and electronic device thereof
JP6123143B1 (ja) 特許要件適否予測装置および特許要件適否予測プログラム
US11249645B2 (en) Application management method, storage medium, and electronic apparatus
US20180165581A1 (en) Electronic apparatus, method of providing guide and non-transitory computer readable recording medium
CN109214674B (zh) 客服系统管理方法、客服系统及电子设备
US11481551B2 (en) Device and method for providing recommended words for character input
WO2020098414A1 (zh) 终端数据处理方法、装置及终端
CN110992937B (zh) 语言离线识别方法、终端及可读存储介质
US20170017655A1 (en) Candidate services for an application
US11789980B2 (en) Method, system, and non-transitory computer readable record medium for providing multi profile
CN111738316B (zh) 零样本学习的图像分类方法、装置及电子设备
CN110990598A (zh) 资源检索方法、装置、电子设备及计算机可读存储介质
JP2024500884A (ja) 有意味自動接続スコアを使用した会話型ボットの評価および強化
US20220374909A1 (en) Data-driven taxonomy for annotation resolution
KR20210045837A (ko) 지식 그래프를 업데이트하는 시스템 및 방법
KR102308062B1 (ko) 창업을 위한 정보를 제공하기 위한 전자 장치 및 그 동작 방법

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: 18924247

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 30.03.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 18924247

Country of ref document: EP

Kind code of ref document: A1