Method of information processing and electronic device

Abstract

To provide an information processing method and an electronic device, which uncover a causality in a target object.SOLUTION: A method includes determining a group of target factors for a target object based on an unstructured text set about the target object. Each target factor represents an aspect of the target object. The method further includes: determining a causal-outcome event pair comprising a causal event and an outcome event by analyzing a text in the text set; and determining, based on the causal-outcome event pair, a first causality between a first factor in the group of target factors and a second factor of the target object.SELECTED DRAWING: Figure 8

Description

Technical Field

[0001] Exemplary embodiments of the present disclosure generally relate to the field of computers, and more particularly, to methods and devices for information processing.

Background Art

[0002] Using unstructured text can provide comments on subjects such as products, services, organizations, etc. For example, user comments are often displayed on product purchase pages or service display pages. As another example, questionnaires include open-ended questions for respondents to provide text comments. Such unstructured text contains rich information about the object to be described, and we want to interpret and utilize such information.

Summary of the Invention

Means for Solving the Problems

[0003] In a first aspect of the present disclosure, a method for information processing is provided. The method includes determining a set of target elements of a target object based on an unstructured text set related to the target object, where each target element represents one aspect of the target object, analyzing the text in the text set to determine a causal event pair including a cause event and a result event, and determining a first causal relationship between a first element of the target element set and a second element of the target object based on the causal event pair.

[0004] In a second aspect of the present disclosure, an electronic device is provided. The electronic device includes at least one processing circuit. The at least one processing circuit is configured to determine a set of target elements of a group of target objects, each representing one aspect of a target object, based on an unstructured text set regarding the target object, and each target element is configured to determine a set of target elements regarding the target object by analyzing the text in the text set based on the unstructured text set regarding the target object, determine a causal event pair including a cause event and a result event, and determine a first causal relationship between a first element of the target element set and a second element of the target object based on the causal event pair.

[0005] In some embodiments of the second aspect, the at least one processing circuit determines at least one reference sentence that matches the causal relationship from the text of the text set. And based on the at least one reference sentence, a target sentence reflecting the first causal relationship is determined.

[0006] In some embodiments of the second aspect, the at least one processing circuit determines at least one reference sentence that matches the causal relationship from the text of the text set. And based on the number of the at least one reference sentence, the causal strength of the first causal relationship indicating the degree of influence of the first element on the second element or the degree of influence of the second element on the first element is determined.

[0007] In some embodiments of the second aspect, the second element includes at least one of an element other than the first element in the target element set, a first predefined element of the target object, and an element of interest of the target object.

[0008] In some embodiments of the second aspect, the text set is obtained from answers to open-ended questions in an information collection sheet for the target object, and the information collection sheet includes closed-ended questions for the first predefined element.

[0009] In some embodiments of the second aspect, at least one processing circuit determines a second causal relationship indicating that a second predefined element of the target affects an element of interest of the target, and determines a second target sentence reflecting the second causal relationship based on a text set so as to determine the second causal relationship indicating that the second predefined element of the target affects the element of interest of the target.

[0010] In some embodiments of the second aspect, determining the first causal relationship includes, when it is determined that the cause event is related to the first element and the result event is related to the second element, determining that the first causal relationship is that the first element affects the second element; and when it is determined that the cause event is related to the second element and the result event is related to the first element, determining that the second element affects the first element.

[0011] In some embodiments of the second aspect, for one of the cause event or the result event, at least one processing circuit determines that the event is related to the first element based on at least one determination that the text representing the event includes a word representing the first element and that the text representing the event is semantically consistent with the word representing the first element.

[0012] In some embodiments of the second aspect, determining a pair of causal events includes determining a pair of causal events according to a self-trained natural language processing model based on the text of the text set.

[0013] In some embodiments of the second aspect, at least one processing circuit presents an element representing the first element and an element representing the second element in association with each other.

[0014] In some embodiments of the second aspect, presenting an element representing the first element and an element representing the second element in association with each other includes presenting a first node representing the first element, presenting a second node representing the second element, and presenting an edge connecting the first node and the second node.

[0015] In some embodiments of the second aspect, at least one processing circuit presents a target sentence that reflects a first causal relationship in relation to an edge.

[0016] In some embodiments of the second aspect, the prominence of an edge is related to the causal strength of a first causal relationship indicating the degree of influence of a first element on a second element or the degree of influence of the second element on the first element.

[0017] In a third aspect of the present disclosure, an electronic device is provided. The electronic device includes at least one processing unit and at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit. The instructions, when executed by the at least one processing unit, are a method for causing the electronic device to execute the first aspect.

[0018] In a fourth aspect of the present disclosure, a computer-readable memory is provided. The computer-readable memory stores a computer program executable by a processor to implement the method of the first aspect.

[0019] It should be understood that what is described in the content of the present disclosure is not intended to limit the important features or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will be readily understood from the following description.

Brief Description of the Drawings

[0020] Referring to the following detailed description in relation to the accompanying drawings, the above features and other features, advantages, and aspects of the embodiments of the present disclosure will become more apparent. In the drawings, the same or similar reference numerals represent the same or similar elements.

[0021]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Embodiments for Carrying Out the Invention

[0022] Hereinafter, embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the accompanying drawings, the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described in the text. Rather, it should be understood that these embodiments are provided to more fully and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are used only for illustrative operations and are not used to limit the scope of protection of the present disclosure.

[0023] In the description of the embodiments of the present disclosure, the terms "including" and similar terms should be understood as open-ended, i.e., "including but not limited to". The term "base" should be understood as "at least partially based on". The term "one embodiment" or "this embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions can also be included hereinafter.

[0024] As used herein, the term "circuit" means a hardware circuit and / or a combination of a hardware circuit and software. For example, a circuit may be a combination of analog and / or digital hardware circuits and software / firmware. As another example, a circuit may be any part of a hardware processor having software that includes (one or more) digital signal processors, software, and (one or more) memories that operate together to enable a device to perform various functions. In yet another example, a circuit may be a hardware circuit and / or a processor such as a microprocessor or a part thereof that requires software / firmware for operation but may not have software if not required for operation. As used herein, the term "circuit" also includes only a hardware circuit or a processor, or a part of a hardware circuit or a processor, and implementations of the software and / or firmware associated with them (or them).

[0025] As used herein, the term "model" can learn the association between corresponding inputs and outputs from training data and generate the corresponding output for a given input after training is completed. The generation of the model is based on machine learning techniques. Deep learning is a machine learning algorithm that uses multiple layers of processing units to process the input and provide the corresponding output. In this document, the "model" can also be referred to as a "machine learning model", a "machine learning network", or a "network", and these terms are used interchangeably in this document. The model further includes different types of processing units or networks.

[0026] As described above, the unstructured text regarding the target contains rich information regarding the target. It is expected that such information can be understood and utilized. On the other hand, causal relationships are widely used in various fields to uncover the causal relationships between various elements of the target. Conventionally, it has been possible to obtain the causal relationships between known elements of the target object. However, in the conventional scenario, the implicit elements included in the unstructured text were not uncovered, and the causal relationships between such implicit elements and other elements of the target could not be clarified. As a result, the excavation of causal relationships regarding the target is not comprehensive.

[0027] Embodiments of the present disclosure propose an aspect for information processing. Based on an unstructured text set regarding a target object, a group of target elements of the target object is determined. Each target element represents one aspect of the target object. By analyzing the text within the text set, a causal event pair including a cause event and a result event is determined. Based on the causal event pair, a first causal relationship between a first element of the target element set and a second element of the target object is determined.

[0028] In this scenario, implicit elements are extracted from the text regarding the target object. Further, causal relationships regarding the implicit elements are determined based on the text. In this way, causal relationships between different implicit elements or between implicit elements and known elements are clarified. Therefore, the present invention helps to improve the excavation of causal relationships of the target object, thus promoting the improvement of the target object.

[0029] <Sample environment> FIG. 1 is a schematic diagram of an exemplary environment 100 in which an embodiment of the present disclosure is implemented. In environment 100, computing device 110 receives text set 105 regarding the target object, or computing device 110 extracts text set 105 from the original data. Text set 105 includes a plurality of texts 101-1, 101-2,... which are also collectively or individually referred to as text 101. The target object includes tangible objects, intangible objects, and combinations thereof. For example, the target object may be a product such as a household item or food. As another example, the target object may be a service such as a cloud computing service or a cloud storage service. As another example, the target may be an entity that provides services or articles such as an airline, a restaurant, or a hotel. The target object may also specify an organizational structure such as a company.

[0030] Text 101 may be a description of the target by the target user. Text 101 includes emotional sentences such as "Apples are delicious" and "Apples are not delicious". Text 101 includes non-emotional sentences such as "I ate an apple" for example. Text 101 may be an evaluation, comments, reviews, assessment, advice, impressions, etc. of the target. Text 101 includes information on factors that affect the target. Individual texts 101 within text set 105 are provided by different users or by the same user at different times.

[0031] In some embodiments, Text 101 may be a user evaluation on the display page of the target. For example, the display page may result from a shopping app (app), a service-providing app, a review app, etc.

[0032] In some embodiments, as shown in FIG. 1, Text 101 may be generated from an information collection sheet 150 for the target. As used herein, an "information collection sheet" is for collecting descriptions (such as evaluations, impressions, etc.) regarding the target and may be, for example, an electronic questionnaire, comments, etc. Information collection sheet 150 includes open-ended questions regarding the target. Computing device 110 collects answers to the open-ended questions in the information collection sheet as Text 101.

[0033] FIG. 2 shows an example of the information collection sheet 150. In this example, the information collection sheet 150 for a certain flight includes open-ended questions 230. The user can provide evaluations of the flight, etc. through the text box. The answer set 250 of the information collection sheet 150 is shown in tabular form. Each row of the answer set 250 represents an answer record from the same respondent. In each answer record, column 258 is the answer to the open-ended question 230. The text 101 may be the text within column 258.

[0034] Continuing to refer to FIG. 1, the computing device 110 determines a group of target elements 102-1, 102-2,... of the target object, which is also collectively referred to as the set of target elements 102 or simply called the target elements 102, based on the text set 105. Since such target elements 102 are determined from unstructured text, they are also called "implicit elements" or "extracted elements". Target elements are regarded as unstructured elements.

[0035] In some embodiments, as shown in FIG. 1, the computing device 110 also receives or determines a group of predefined elements 103-1, 103-2,... of the target object, which is also collectively referred to as the set of predefined elements 103 or separately called the predefined elements 103. The term "predefined element" as used herein means an element whose measurement criterion has a predetermined option (e.g., a predetermined numerical value, category, star, etc.). For a predefined element, the user selects one option from the predefined options to evaluate or describe the target object from the perspective of that predefined element. Predefined elements are quantitative and highly organized. The description (e.g., evaluation, assessment) of predefined elements is not open-ended and must conform to an architecture with predefined options. Therefore, predefined elements are also regarded as structured elements.

[0036] The predefined elements can also include numerical elements or categorical elements. The predetermined options of the numerical elements include predetermined numerical values, stars, etc. The predetermined options of the categorical elements include predetermined classes such as the class of the cabin, etc.

[0037] In some embodiments, as shown in FIG. 1, the computing device 110 can also receive or determine elements of interest for the target. As used herein, "element of interest" means an aspect that the target is particularly focused on. Elements of interest include the performance, service, function, overall performance, overall evaluation, or satisfaction of the target.

[0038] The set of elements 130 of the target includes the target element 102 and any predefined elements 103 and elements of interest 104. In this text, the target element, predefined element, and element of interest are collectively or individually also referred to as "elements".

[0039] In some embodiments, as shown in FIG. 1, the predefined element 103 may be obtained from the information collection sheet 150. The information collection sheet 150 includes closed-ended questions regarding the predefined element 103. A "closed-ended question" is a question whose answer is selected from a predefined set of options. In the example of FIG. 2, the information collection sheet 150 includes a closed-ended question 210-1 regarding the predefined element "Seat comfort", a closed-ended question 210-2 regarding the predefined element "Cabin service", a closed-ended question 210-3 regarding the predefined element "Food beverage", a closed-ended question 210-4 regarding the predefined element "Entertainment", a closed-ended question 210-5 regarding the predefined element "Ground service", and a closed-ended question 210-6 regarding the predefined element "Value for money". The closed-ended questions 210-1 to 210-6 are collectively referred to as the closed-ended questions 210 or are called individually. Each closed-ended question 210 has five scores for the user to select. In the answer set 250, columns 252 to 257 are the user's answers to the closed-ended questions 210-1 to 210-6, respectively.

[0040] The element of interest 104 is also obtained from the information collection sheet 150. In the example of FIG. 2, the information collection sheet 150 includes a closed-ended question 220 regarding the overall satisfaction of the target. In the answer set 250, column 251 is the user's answer to the closed-ended question 220. Therefore, in this example, the element of interest of the target is "Overall satisfaction".

[0041] In the example of FIG. 2, the target is air freight. As another example, the target may be a certain product. The predefined elements include elements such as the appearance, quality, price, features, etc. of the product. The element of interest may be, for example, the brand value of the product. As yet another example, the target may be a company. The predefined elements include the salary of employees, work location, working hours, promotion path, etc. Examples of elements of interest include employee satisfaction and employees' intention to stay, etc.

[0042] Continue to refer to FIG. 1. The target element 102 is an element implicitly included in the unstructured text. It is desirable to clarify the causal relationship between the target element 102 and other elements of the target. For example, it is desirable to determine whether the target element 102 has a causal relationship with other target elements, predefined element 103, or element of interest 104.

[0043] For this target, the computing device 110 determines the causal relationship between the target element 102 and other elements based on the text in the text set 105. In some embodiments, a causal relationship model for processing text input is utilized. The model determines the target element 102 and the causal relationships associated with the target element 102 based on the text in the text set 105.

[0044] In some embodiments, the computing device 110 determines one or more causal event pairs 120-1, 120-2,... that are also collectively referred to as causal event pairs 120 or individually called causal event pairs 120 by analyzing the text 101 in the file set 105. Each pair of causal events includes a cause event and a corresponding result event. For example, the causal event pair 120-1 includes the cause event 121-1 and the result event 122-1, and the causal event pair 120-2 includes the cause event 121-2 and the result event 122-2. The cause events 121-1, 121-2,... are also collectively referred to as the cause event 121 or individually called, and the result events 122-1, 122-2,... are also collectively referred to as the result event 122 or individually called.

[0045] Computing device 110 generates a causal relationship set 160 by matching a cause event 121 and an effect event 122 with elements within an element set 130. The causal relationship set 160 includes at least the causal relationship between a certain target element 102 and another element of the target object (for example, another target element, a predefined element, an element of interest). In some embodiments, the causal relationship set 160 also includes the causal relationship between a predefined element 103 and an element of interest 104. The computing device 110 obtains such a causal relationship in any suitable way, such as receiving it from other devices or determining it by itself using a suitable method. Embodiments of the present disclosure are not limited in this regard.

[0046] In environment 100, the computing device 110 may be any type of computing device, including a terminal device or a service end device. The terminal device may be any type of mobile terminal, fixed terminal or portable terminal, including a mobile handset, a desktop computer, a laptop computer, a netbook computer, a tablet computer, a media computer, a multimedia tablet, a personal communication system (PCS) device, a personal navigation device, a personal digital assistant (PDA), an audio / video player, a digital camera / video camera, a positioning device, a television receiver, a radio broadcast receiver, an e-book device, a game device, or any combination of the above including components and peripheral devices of these devices. The service-side device includes, for example, a computing system / server such as a mainframe, an edge computing node, a computing device within a cloud environment.

[0047] It should be understood that the structure and functions of the environment 100 are described only for exemplary targets and do not imply any limitation to the scope of the present disclosure. Although one computing device 110 is separately shown in FIG. 1, in some embodiments, the various operations described herein may be implemented by multiple computing devices. Although the target elements and causal relationships are shown separately, the target elements and corresponding causal relationships may be determined by the same model based on the text set 105.

[0048] Furthermore, the information collection sheet shown in FIG. 2 is merely exemplary and is not intended to limit the scope of the present disclosure. The open questions, closed questions, and their numbers shown in FIG. 2 are merely exemplary. In embodiments of the present disclosure, the information collection sheet may have any suitable number of open questions and closed questions. Further, although English is mentioned as an example, embodiments of the present disclosure may be used to process text and information collection sheets in any language.

[0049] <Extraction of Target Elements> As described above with reference to FIG. 1, in order to dig out potential causal relationships related to the target object, it is necessary to extract implicit elements from unstructured text. An example of the process is described below. FIG. 3 is a flowchart of a process 300 for determining target elements according to some embodiments of the present disclosure. The process 300 is implemented in the computing device 110. For ease of explanation, the process 300 will be described with reference to FIG. 1.

[0050] At block 310, the computing device 110 extracts a plurality of keywords from the unstructured text set 105 for the target subject. The extracted keywords may have any suitable number of word splits. The keywords may include one word keywords, such as "flight", "seat", "service", and two word keywords, such as "cabin crew", "flight attendant", and may utilize any suitable keyword extraction algorithm, such as, but not limited to, using TF-IDF, KP-Miner, SBKE, RAKE, TextRank, YAKE, KeyBERT, and the like.

[0051] In some embodiments, before applying the keyword extraction algorithm, the text 101 in the text set 105 can be pre-processed, e.g., removing named entities and stop words. Named entities are, e.g., people's names, institution names, place names, etc., and do not describe any aspect of the target subject. For English, stop words are, e.g., "a", "an", "the", "and", etc. For Chinese text, stop words are, e.g., "one", "one", "and", "but", etc. Alternatively, in some embodiments, the text 101 may be pre-processed by the keyword extraction algorithm.

[0052] In some embodiments, a keyword extraction algorithm is used to extract nouns as keywords from the text set 105, which avoids extracting words that cannot describe other attributes of the target object's aspect, thereby effectively reducing the difficulty of subsequent processing.

[0053] In some embodiments, computing device 110 extracts keywords based on the number of occurrences of each word in text set 105 (i.e., word frequency). Specifically, computing device 110 extracts candidate words from text 101 of text set 105. If the number of occurrences of a candidate word in text set 105 is greater than a threshold number of times, then that candidate word is determined to be one of the keywords. If the number of occurrences of a candidate word in text set 105 is less than the threshold number of times, the candidate word is deleted.

[0054] For example, using a keyword extraction algorithm, candidate words are extracted from column 258 of each answer record. For each extracted candidate word, the number of occurrences of the candidate word in the entire text set 105 is calculated. Then, candidate words with a number of occurrences greater than the threshold number of times are determined to be keywords, and candidate words with a number of occurrences less than the threshold number of times are deleted. In such embodiments, by filtering the pre-extracted candidate words, it is avoided that unimportant words interfere with the determination of the target element.

[0055] Alternatively, in some embodiments, computing device 110 extracts keywords based on the meaning of text 101 within text set 105. For example, by semantic analysis, sentences with sentiment are determined, and nouns related to the sentiment in such sentences are used as keywords.

[0056] In block 320, computing device 110 groups at least some of the plurality of keywords based on the meaning of the plurality of keywords. In some embodiments, all keywords are grouped. In some embodiments, the keywords may be filtered based on the result of a preliminary grouping, and the filtered keywords may be grouped.

[0057] Computing device 110 utilizes a cluster to group a plurality of extracted keywords. Therefore, a word vector representing the meaning is generated for each keyword. The word vector is generated using any suitable method such as word2vector, GloVe, etc. Embodiments of the present disclosure are not limited in this regard.

[0058] A plurality of keywords can be clustered based on the word vectors to determine a plurality of clusters, and each cluster includes at least one keyword. The clustering algorithm divides these keywords into independent and non-overlapping clusters based on the semantic similarity of the keywords. Any suitable clustering algorithm is adopted, for example, K-Means, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), Gaussian mixture model, etc.

[0059] In some embodiments, the keywords may be filtered based on the quality of each cluster. The quality of a cluster represents how semantically concentrated the keywords within that cluster are. For example, the sum of the squared distances of the keywords within the cluster is used as the quality of the cluster. Alternatively or additionally, the Silhouette coefficient can also be used as the quality of the cluster.

[0060] Determine the quality of each cluster obtained by clustering. In some embodiments, remove keywords within a cluster whose quality is lower than a threshold quality, and determine the remaining keywords. The remaining keywords are grouped based on the meaning of the remaining keywords. For example, cluster the remaining keywords. Consider the keywords within the same obtained cluster as a keyword set. Alternatively, in some embodiments, remove clusters having a quality lower than the threshold quality and retain other clusters having a quality higher than the threshold quality. In the reserved clusters, the keywords within the same cluster are treated as a keyword set. In such embodiments, there is no need to regroup the remaining keywords.

[0061] Figure 4 is a diagram showing an example of keyword grouping. The grouped result is obtained by processing the text within column 258 in answer set 250. In Figure 4, keyword group 410, keyword group 420, keyword group 430, keyword group 440, keyword group 460, and keyword group 470 are determined by clustering. Each keyword group includes one or more keywords.

[0062] Continuing to refer to Figure 3. In block 330, computing device 110 determines target element 102 corresponding to the keyword set based on the result of grouping. Target element 102 represents one aspect of the target object. Since the same keyword set has a similar meaning, it represents the same aspect of the target object. Thus, the keyword set corresponds to one target element 102.

[0063] The name or identification of the target element 102 corresponding to the keyword set is determined based on the keyword set. As an example, any of the keyword sets is used to represent the corresponding target element. As another example, the center of a cluster consisting of the keyword set is determined, and the target element corresponding to the keyword having the semantic feature closest to the center is represented. As a further example, the target element is represented by an aspect of the target object (e.g., service or performance) described by the keyword set.

[0064] In the example of FIG. 4, the target element corresponding to the keyword group 410 is "Tv service TV (television service)". The target element corresponding to the keyword group 420 is "boarding procedure (boarding procedure)". The target element corresponding to the keyword group 430 is "luggage service (luggage service)". The target element corresponding to the keyword group 440 is "movie service (movie service)". The target element corresponding to the keyword group 460 is "time delay (time delay)". The target element corresponding to the keyword group 470 is "legroom (legroom)".

[0065] In some embodiments, one or more keyword sets identical or similar to the predefined elements are removed. In this case, the computing device 110 determines the target elements corresponding to the keyword groups that have not been removed. For example, for each keyword set, the computing device 110 determines whether the keyword set is semantically similar to the predefined elements of the target object. If the keyword set is not similar to any of the semantically defined elements, the target element is determined based on the keyword set. If the keyword set is semantically similar to the predefined element, the keyword set is deleted.

[0066] As an example, by processing the text in column 258, a set of keywords "food", "meal", "drink", "snack" is obtained. The keyword set is semantically similar to the predefined element "food and beverage" in FIG. 2. Therefore, the keyword set is removed without determining the corresponding target element.

[0067] By the above process 300, target elements are extracted from open text comments or comments. In this way, analyzing the information contained in such unstructured text discovers new elements that affect the target. It should be understood that process 300 is exemplary. In embodiments of the present disclosure, any suitable method is used to extract implicit elements from the text.

[0068] <Causal relationship extraction> As described with reference to FIG. 1, in addition to the target element 102, the computing device 110 extracts a causal event pair 120 including a cause event 121 and a result event 122 from the text set 105. Further, the computing device 110 matches the cause event and the result event with the elements in the element set 130 to determine the causal relationship between the two elements.

[0069] FIG. 5 shows a schematic diagram of an architecture 500 for causal relationship extraction according to some embodiments of the present disclosure. As shown in FIG. 5, the language conversion module 501 converts the text 101 into text in the target language. The language conversion module 501 may be implemented in any suitable way. The language conversion module 501 is realized, for example, by translating an application programming interface (API).

[0070] If the text set 105 contains texts in different languages, or if the text 101 is different from the target language, the language conversion module 501 is used to convert the text 101 into the same target language. It should be understood that in some embodiments, the language conversion module 501 may not be used. For example, when the text 101 is all in the target language, or when the event pair extraction module 503 can handle texts in multiple languages, the language conversion module 501 may not be used.

[0071] The text 101 having the target language can be supplied to the event pair extraction module 502. The event pair extraction module 502 determines a causal event pair 510 including a cause event 511 and a result event 512 by analyzing the text 101. It is understood that the causal event pair 510 is an example of the causal event pair 120 shown in FIG. 1. Further, although only one causal event pair 510 is shown, the event pair extraction module 502 extracts any number of causal event pairs.

[0072] The event pair extraction module 502 uses any suitable method to determine the causal event pair 510. For example, a rule template is used to extract the causal event pair 510 in the text 101. As another example, the causal event pair 510 is extracted using a model modeled as a sequence marking task. In this example, the text 101 is represented as a sequence of words or characters, a sequence of labels of the same length is output by the model, and the positions of the cause event and the result event in the text are identified by that sequence of labels. Such a model is trained end-to-end.

[0073] In some embodiments, the event pair extraction module 502 determines causal event pairs 510 based on the text 101 in the text set 105 according to a self-training natural language processing (NLP) model. The NLP model is configured to extract causal event pairs from natural language text. This NLP model is trained simultaneously on a small amount of tagged data and a large amount of untagged data. The training of this NLP model does not rely on a large amount of tagged data and fully utilizes the information from the untagged data to improve the performance of the model.

[0074] As an example, the NLP model includes a language characterization model, a conditional airport model, and a noise adaptation layer. The language characterization model is used to generate word vectors of the text 101 and is implemented with any suitable network structure. For example, this language characterization model may be a Bidirectional Encoder Representations from Transformers (BERT) model. The conditional airport model, together with the noise adaptation layer, is used for labeling unlabeled data. The noise adaptation layer is used to generate a noise matrix for each word from the word vector of that word.

[0075] Compared with task models such as template matching and sequence marking, the self-training NLP model has high accuracy and high generalization ability. By using such a model for causal event pair extraction, causal event pairs can be extracted more accurately and comprehensively.

[0076] The element matching module 503 determines the causal relationship 520 between the element 521 and the element 522 based on the causal event pair 510. In the example of FIG. 5, the causal relationship 520 is that the element 521 affects the element 522. At least one of the element 521 and the element 522 is the target element 102. When the element matching module 503 determines that the cause event 511 is involved in the element 521 and the result event 512 is involved in the element 522, the causal relationship 520 is determined to be that the element 521 affects the element 522.

[0077] In some embodiments, whether it is related to an event element may be determined by exact matching. If a word representing an element is included in the text representing an event, it is determined that the event is related to that element. For example, when the text representing the cause event 511 includes the word representing the element 521, the element matching module 503 determines that the cause event 511 is related to the element 521. When the text representing the result event 512 includes the word representing the element 522, the element matching module 503 determines that the result event 512 includes the element 522.

[0078] The text representing an event is obtained from the text set 105. For example, when executing event pair extraction, the event pair extraction module 502 determines the text representing the causal event pair, and determines the text representing the cause event and the text representing the result event therefrom. The word representing an element may be the name of the element or a part of the name. For example, in the case of the target element "leg room", the word representing the element may be "leg room" or "leg".

[0079] Alternatively or additionally, the word representing an element may be a word that is semantically close to that element. In the case of the target element 102, the word representing the target element 102 is a keyword within the keyword group used to determine the target element in the process 300. For example, in the case of the target element "luggage service", the words representing the element may be the words "bag", "luggage", "baggage", and "hand luggage" within the keyword group 430. In the case of the predefined element 103 or the element of interest 104, the words that are semantically close to the element are determined in any suitable way. Embodiments of the present disclosure are not limited in this regard.

[0080] Alternatively or additionally, in some embodiments, a fuzzy matching method is used to determine whether an event element is related. When the text representing an event and the word representing an element are semantically consistent, it is determined that the event is related to the element. For example, if the text representing the cause event 511 is consistent with the meaning represented by the element 521, the element matching module 503 determines that the cause event 511 is related to the element 521. If the text representing the result event 512 is consistent with the meaning represented by the element 522, the element matching module 503 determines that the result event 512 is related to the element 522.

[0081] The element matching module 503 uses any suitable fuzzy matching method. For example, the text representing an event and the word representing an element are input into a language characterization model (e.g., BERT) for encoding to generate an embedded vector of the event and an embedded vector of the element. If the difference between the two embedded vectors is smaller than the threshold difference, it is determined that the text representing the event is consistent with the meaning represented by the element. Therefore, it is determined that the event is involved with the element. Also, the Jaccard similarity is utilized to determine whether the text representing the event and the word representing the element are semantically consistent.

[0082] In the above, the operation example of the element matching module 503 has been described. To more clearly understand that the event pair extraction according to an embodiment of the present disclosure matches the element, some examples are described below. FIG. 6 shows an example of a causal relationship set 160 according to some embodiments of the present disclosure.

[0083] Generally, the causal relationship set 160 relates to the interest element "Overall satisfaction", the predefined element "Value for money", the predefined element "Entertainment", the predefined element "Seat comfort", the predefined element "Cabin service", the predefined element "Food beverage", and the predefined element "Ground service", as described with reference to Figure 2. As mentioned above, the causal relationships 601, 602, 603, 604, 605, and 606 between each predefined element and the interest element "Overall satisfaction" may be obtained using any suitable causal relationship model.

[0084] The causal relationship set 160 also relates to the extracted target elements "Movie service", "Leg room", "Time delay", "Boarding service" and "Luggage service" as described with reference to Fig. 4. Through causal time extraction and element matching, we determine causal relationships 611, 612, 613, 614, 615 for these target elements.

[0085] As an example, by analyzing the text "Extra luggage weight leads to that we are unable to check in together" in the text set 105, a causal event pair including a cause event "extra luggage weight" and an effect event "unable to check in together" is identified. The cause event "extra luggage weight" matches the target element "Luggage service", and the effect event "unable to check in together" matches the target element "Boarding service". Thus, a causal relationship 615 is determined in which the target element "Luggage service" affects the target element "Boarding service".

[0086] As another example, by analyzing the text "The 3-3-3 seats and plenty of leg room make the seats quite spacious" in text set 105, a causal event pair including the causal event "plenty of leg room" and the result event "the seats quite spacious" is identified. The causal event "plenty of leg room" matches the target element "Leg room", and the result event "the seats quite spacious" matches the predefined element "Seat comfort". Therefore, a causal relationship 612 in which the target element "Leg room" affects the predefined element "Seat comfort" is determined.

[0087] Similarly, by analyzing the text "Chaotic boarding process causes a delay" in text set 105, a causal relationship 614 in which the target element "Boarding service" affects the target element "Time delay" is determined. By analyzing the text "Delay makes me unsatisfied with the flight" in text set 105, a causal relationship 613 in which the target element "Time delay" affects the interest element "Overall satisfaction" is determined.

[0088] Depending on the text within text set 105, in some cases, multiple causal relationships are determined based on one causal event pair. When the cause event or the result event of the causal event pair involves two or more elements, the multiple causal relationships are determined based on the causal event pair. For example, by analyzing the text "Newly released tv shows and movies improved entertainment" within text set 105, the cause event "Newly released tv shows and movies" and the result event "entertainment" are identified. Since two elements, the target elements "Movie service" and "TV service", are involved in the cause event, two causal relationships are identified. Specifically, a causal relationship 611 in which the target element "Movie service" affects the predefined element "Entertainment" and a causal relationship 616 in which the target element "TV service" affects the predefined element "Entertainment" are determined.

[0089] The set of causal relationships 160 shown in FIG. 6 is merely exemplary and is not intended to limit the scope of the present disclosure. The set of causal relationships 160 includes any number of elements. In some embodiments, the causal relationships 160 relate only to target elements without relating to predefined elements and elements of interest. Also, in the above example described with reference to FIG. 6, the causal event pair is determined based on one sentence, but this is merely exemplary. The same causal event pair is extracted from multiple different sentences.

[0090] The above described an example of causal event pair extraction and element matching. In an embodiment of the present disclosure, implicit elements are mined from text, causal event pairs are extracted from the text, and based on the causal event pairs that match the target element, a causal relationship related to the target element is determined. On the other hand, in this way, the causal relationship regarding the target object is dug up and complemented. On the other hand, since the implicit elements and causal event pairs are generated from the text regarding the target object, a more accurate causal relationship can be obtained.

[0091] Continuing to refer to FIG. 5. In some embodiments, the computing device 110 can also determine the causal strength 531 of the causal relationship 520. The causal strength 531 indicates the degree of influence that the element 521 exerts on the element 522. The causal strength 531 of the causal relationship 520 is determined based on the number of reference sentences that match the causal relationship 520.

[0092] The reference sentence that matches the causal relationship 520 may be a sentence that semantically expresses the causal relationship 520. For example, the reference sentence includes the sentence from which the corresponding causal event pair 510 was extracted. For example, in the case of the causal relationship 611 shown in FIG. 6, the reference sentence includes "Newly released tv shows and movies improved entertainment". For the causal relationship 612, the reference sentence includes "The 3-3-3 seats and plenty of leg room make the seats quite spacious".

[0093] In some embodiments, computing device 110 can also determine a target sentence 532 that reflects causal relationship 520. Computing device 110 further determines target sentence 532 based on at least one reference sentence that matches causal relationship 520. In some embodiments, one or more of the reference sentences are selected as target sentence 532. For example, use the sentence "Chaotic boarding process causes a delay" as the target sentence for causal relationship 614. Alternatively or additionally, appropriate natural language processing methods are used to fuse the reference sentences into target sentence 532.

[0094] Showing causal relationships individually, the causal relationships can be interpreted using target sentences that are not intuitive and easy to understand, and one can intuitively feel how element 521 affects element 522. The target sentence also serves as evidence of the causal relationship. This allows the elements to be adjusted to improve the target.

[0095] In some embodiments, computing device 110 determines a target sentence that reflects the causal relationship between predefined element 103 and interested element 104. A reference sentence that matches this causal relationship is determined from the text of text set 105, and the target sentence is determined based on the reference sentence. For example, one or more reference sentences are specified as the target sentence, and multiple reference sentences are combined into the target sentence.

[0096] As an example, a sentence containing both the predefined element 103 and the element of interest 104 is extracted from the text set 105. If the causal trigger words in the sentence (e.g., "because", "cause", "therefore", etc.) indicate that the predefined element 103 is involved in the cause event and the element of interest 104 is involved in the result event, the sentence is determined to be a reference sentence. As shown in FIG. 6, for the causal relationship 603 between the predefined element "Seat comfort" and the element of interest "Overall satisfaction", an example of the target sentence is "Comfortable seating and the latest entertainment make this flight one of the highlights of our trip".

[0097] Similar to the determination of the causal relationship, there may be sentences that reflect multiple causal relationships. For example, there are two elements in the cause event of the sentence "Comfortable seating and the latest entertainment make this flight one of the highlights of our trip". Thus, it functions as both the target sentence of the causal relationship 602 and the target sentence of the causal relationship 603.

[0098] Individually presenting the causal relationship between the predefined element 103 and the element of interest 104 may not be intuitive and easy to understand. In this way, the causal relationship between the predefined element 103 and the element of interest 104 is explained.

[0099] It should be understood that the various modules in the architecture shown in FIG. 5 are merely exemplary and are not intended to limit the scope of the present disclosure. The operations and functions described with reference to FIG. 5 may be implemented by the same module or model.

[0100] Also, although described separately, the determination of target elements and the extraction of causal relationships are realized by the same module or model. Such a module or model is configured to determine target elements and causal relationships related to the target elements based on a text set. In other words, a natural language text set is input into such a module or model to obtain a set of causal relationships including target elements and causal relationships related to the target elements.

[0101] <Presentation of the set of causal relationships> The set of causal relationships 160 may be presented on the computing device 110 or other suitable display device. For this purpose, it is presented in association with the element representing the element 521 and the element representing the element 522. In some embodiments, the causal relationships within the set of causal relationships 160 may be presented in a table. For example, two related cells in the table represent the element 521 and the element 522, respectively.

[0102] In some embodiments, the causal relationships within the set of causal relationships 160 may be presented in the form of an image. FIG. 7 shows an example of an image 700 representing causal relationships according to some embodiments of the present disclosure. The image 700 is used to represent the set of causal relationships 160 of FIG. 6.

[0103] Each node in the image 700 represents each element of the set of causal relationships. For example, the node 701 represents the element of interest "Overall satisfaction", the node 702 represents the predefined element "Sear comfort", and the node 703 represents the target element "Leg room". An edge with a direction connecting two nodes represents the causal relationship between the corresponding elements. For example, the edge 711 directed from the node 702 to the node 701 indicates that the predefined element "Sear comfort" affects the element of interest "Overall satisfaction". As another example, the edge 712 directed from the node 703 to the node 702 indicates that the target element "Leg room" affects the predefined element "Sear comfort".

[0104] In some embodiments, the prominence of an edge (e.g., color, thickness, etc.) is related to the strength of the corresponding causal relationship. For example, edge 713 is thicker than edge 714, which means that the causal relationship between the target element "Boarding service" and the target element "Time delay" is stronger than the causal relationship between the target element "Luggage service" and the target element "Boarding service". As another example, edge 712 is thicker than edge 713, which means that the causal relationship between the target element "Leg room" and the predefined element "Seat comfort" is stronger than the causal relationship between the target element "Boarding service" and the target element "Time delay". In such embodiments, the strength of the causal relationship is visually indicated.

[0105] In some embodiments, target sentences reflecting corresponding causal relationships may be displayed in relation to the edges in image 700. FIG. 7 shows text boxes 720-1, 720-2, 720-3, 720-4, 720-5, and 720-6, which are also referred to as text boxes 720 either individually or collectively. A target sentence reflecting the causal relationship regarding the target element is presented in text box 720. For example, a target sentence such as "Chaotic boarding process causes a delay", which reflects the causal relationship between "Boarding service" and "Time delay", is presented in text box 720-1. As another example, "Newly released tv shows and movies improved entertainment" is displayed in text box 720-5 and text box 720-6. As another example, a target element reflecting the causal relationship between the target element "Leg room" and the predefined element "Seat comfort", such as "The 3-3-3 seats and plenty of leg room make the seats quite spacious", is presented in text box 720-4.

[0106] Regarding the causal relationship between the predefined element and the element of interest, if a target sentence reflecting such a causal relationship exists within text set 105, such a target sentence is presented. FIG. 7 also shows text boxes 730-1 and 730-2, which are also referred to as text boxes 730 either individually or collectively. For example, the sentence "Comfortable seating and the latest entertainment make this flight one of the highlights of our trip" can be expressed in text box 730-1 and text box 730-2.

[0107] The rendering of the text box 720 may be dynamic. For example, the text box 720 is presented in response to detecting a click or selection on the edge 713. In such an embodiment, by displaying the target sentence, the causal relationship relevant to the user interested in the target object is intuitively understood. This helps the user to identify the measures to be improved.

[0108] <Exemplary Process> Figure 8 is a flowchart of a process 800 of information processing according to some embodiments of the present disclosure. The process 800 is implemented on the computing device 110. For ease of explanation, the process 800 will be described with reference to FIG. 1.

[0109] In block 810, the computing device 110 determines a set of target elements 102 of one group of the target object based on the set of unstructured text 105 regarding the target object, and each target element 102 represents an aspect of the target object. In block 810, the computing device 110 determines the target elements 102 using any suitable method. For example, the computing device 110 executes the process 300 described above to extract the target elements 102 from the text set 105.

[0110] In block 820, the computing device 110 determines causal event pairs including cause events and result events by analyzing the text within the text set 105. In some embodiments, the computing device 110 determines the causal event pairs according to a self-trained natural language processing model based on the text within the text set 101.

[0111] In block 830, computing device 110 determines a first causal relationship between a first element among one group of target elements and a second element of a target object based on causal event pairs. The first element may be any target element, and the second element may be at least one of another target element, a predefined element of the target object, or an element of interest.

[0112] In some embodiments, text set 105 is obtained from answers to open-ended questions within information collection sheet 150 for a target object, and information collection sheet 150 includes closed-ended questions for predefined elements.

[0113] In some embodiments, if it is determined that a cause event is related to the first element and a result event is related to the second element, the first causal relationship determines that the first element affects the second element. Alternatively, if it is determined that a cause event is related to the second element and a result event is related to the first element, the first causal relationship determines that the second element affects the first element.

[0114] In some embodiments, for one of a cause event or a result event, it is determined that the event includes the first element based on at least one determination that the text representing the event includes a word representing the first element or that the text representing the event is semantically consistent with a word representing the first element.

[0115] In some embodiments, process 800 further includes additional steps. Computing device 110 determines at least one reference sentence that matches the relationship of the causal event pair from the text of text set 105, and determines a target sentence that reflects the first causal relationship based on the at least one reference sentence.

[0116] In some embodiments, computing device 110 determines at least one reference sentence that matches a pair of causally related events from the text of text set 105, and determines the causal strength of a first causal relationship based on the number of at least one reference sentence. Causal strength indicates the degree of influence of a first element on a second element, or the degree of influence of the second element on the first element.

[0117] In some embodiments, based on a text set, computing device 110 determines a second causal relationship indicating that a second predefined element of a target object affects an element of interest of the target object, and determines a second target sentence that reflects the second causal relationship.

[0118] In some embodiments, computing device 110 presents an element representing a first element in association with an element representing a second element.

[0119] In some embodiments, to present an element representing a first element in association with an element representing a second element, computing device 110 presents a first node representing the first element and a second node representing the second element, and presents an edge connecting the first node and the second node.

[0120] In some embodiments, computing device 110 presents a target sentence that reflects the first causal relationship in relation to the edge.

[0121] In some embodiments, the prominence of the edge may be related to the causal strength of the first causal relationship. Causal strength indicates the degree of influence of a first element on a second element, or the degree of influence of the second element on the first element.

[0122] <Sample Device> FIG. 9 is a block diagram showing a computing device 900 in which one or more embodiments of the present disclosure may be implemented. It should be understood that the computing device 900 shown in FIG. 9 is merely an example and should not constitute any limitation on the functions and scope of the embodiments described in the text. The computing device 900 shown in FIG. 9 is used to implement the computing device 110 of FIG. 1.

[0123] As shown in FIG. 9, the computing device 900 is in the form of a general-purpose computing device. The components of the computing device 900 include, but are not limited to, one or more processors or processing units 910, a memory 920, a storage device 930, one or more communication units 940, one or more input devices 950, and one or more output devices 960. The processing unit 910 may be an actual processor or a virtual processor and executes various processes according to a program stored in the memory 920. In a multiprocessor system, multiple processing units execute computer-executable instructions in parallel to improve the parallel processing ability of the computing device 900.

[0124] Computing device 900 generally includes multiple computer memories. Such media may be any accessible media accessible by computing device 900, including but not limited to volatile media and non-volatile media, removable media and non-removable media. Memory 920 may be volatile memory (e.g., registers, caches, random access memory (RAM)), non-volatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. Storage device 930 can be a removable or non-removable medium and includes machine-readable memory such as a flash drive, magnetic disk, or any other medium that can be used to store information and / or data (e.g., training data for training) and can be accessed within computing device 900.

[0125] Computing device 900 further includes removable / non-removable volatile / non-volatile memory. Although not shown in FIG. 9, it includes a disk drive for reading from or writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable non-volatile optical disk. In these cases, each drive may be connected to a bus (not shown) by one or more data media interfaces. Memory 920 includes a computer program product 925 having one or more program modules configured to execute various methods or operations of various embodiments of the present disclosure.

[0126] The communication unit 940 enables communication with other computing devices via a communication medium. Further, the functions of the components of the computing device 900 may be implemented on a single computing cluster or multiple computer machines that can communicate via a communication connection. Thus, the computing device 900 operates in a network environment using a logical connection with one or more other servers, network personal computers (PCs), or other network nodes.

[0127] The input device 950 is one or more input devices such as a mouse, keyboard, trackball, etc. The output device 960 is one or more output devices such as a display, speaker, printer, etc. The computing device 900 can also communicate, as needed, via the communication unit 940 with one or more external devices (not shown) such as a storage device, display device, etc., communicate with one or more devices that enable a user to interact with the computing device 900, or communicate with any device (e.g., network card, modem, etc.) that enables the computing device 900 to communicate with one or more other computing devices. Such communication is performed via an input / output (I / O) interface (not shown).

[0128] According to an exemplary implementation of the present disclosure, there is provided a computer-readable memory storing computer-executable instructions executed by a processor to implement the methods described above. According to an exemplary implementation of the present disclosure, there is also provided a computer program product tangibly stored on a non-transitory computer-readable memory and including computer-executable instructions executed by a processor to implement the methods described above.

[0129] Various aspects of the present disclosure are described herein with reference to flowcharts and / or block diagrams of methods, devices, apparatus, and computer program products implemented in accordance with the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, may be implemented by computer-readable program instructions.

[0130] These computer-readable program instructions are provided to a processing unit of a general purpose computer, a special purpose computer, or other programmable data processing device, such that when the instructions are executed via the processing unit of the computer or other programmable data processing device, they manufacture a machine capable of realizing the functions / operations defined in one or more blocks in the flowchart and / or block diagram. These computer-readable program instructions may also be stored in a computer-readable memory that causes a computer, programmable data processing device, and / or other device to operate in a particular manner, such that the computer-readable memory storing the instructions includes an article of manufacture of instructions implementing various aspects of the functions / operations defined in one or more blocks in the flowchart and / or block diagram.

[0131] The computer-readable program instructions are loaded onto a computer, other programmable data processing device, or other device, and a series of operational steps executed on the computer, other programmable data processing device, or other device are performed to generate a computer-implemented process such that the instructions executed on the computer, other programmable data processing device, or other device implement the functions / operations defined in one or more blocks in the flowchart and / or block diagram.

[0132] Flowcharts and block diagrams illustrate the architecture, functionality, and operation of possible embodiments of systems, methods, and computer program products, in accordance with various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram represents a module, a program segment, or a portion of code that contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two consecutive blocks may actually be executed substantially in parallel, or may sometimes be executed in the reverse order, depending upon the functionality involved. Each block in the block diagrams and / or flowcharts, and / or combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified function or operation, or may be implemented by a combination of dedicated hardware and computer instructions.

[0133] Although embodiments of the present disclosure have been described above, the above description is illustrative and not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The choice of terms used in this text is intended to best explain each embodiment disclosed in this text, the principles of each embodiment disclosed in this text, the actual application, or the improvement of the technology in the market, or to be understandable to other ordinary technicians in the art.

Claims

1. A method for information processing executed by a computing device, comprising: determining target elements of one group of the target objects based on an unstructured text set regarding a target object of interest to a user; determining a causal event pair including a cause event and a result event by analyzing the text in the unstructured text set; determining a first causal relationship between a first element among the target elements of the one group and a second element of the target object based on the causal event pair; wherein the unstructured text set can provide comments on the target object; the determining of the target elements includes using a keyword extraction algorithm to extract a plurality of keywords from the unstructured text set; grouping at least a part of the plurality of keywords based on the meanings of the plurality of keywords; determining the target elements corresponding to the keyword(s) based on the result of the grouping; A method for information processing.

2. The second element includes at least one of an element other than the first element among the target elements of the one group, a first predefined element of the target object, and an element of interest of the target object; the first predefined element is an element having a measurement criterion with a predetermined option; the element of interest is an aspect on which the target object is focused; The method according to claim 1.

3. The unstructured text set consists of answers to open-ended questions in an information collection sheet for the target object, the information collection sheet includes closed-ended questions for the first predefined element; The method according to claim 2.

4. Regarding one of the cause event or the result event, the text representing the event includes a word representing the first element; the text representing the event is semantically consistent with the word representing the first element; further comprising being determined to be involved in the first element based on at least one of the above; The method according to claim 1.

5. further comprising that an element representing the first element and an element representing the second element are presented in association; The method according to claim 1.

6. The elements representing the first element and the elements representing the second element presented in association are Present a first node representing the first element and a second node representing the second element, and An edge connecting the first node and the second node, The method according to claim 5, comprising

7. At least one processing circuit, wherein the at least one processing circuit Based on an unstructured text set regarding a target object of interest to a user, determine target elements of one group of the target object, By analyzing the text in the unstructured text set, determine causal event pairs including cause events and result events, Based on the causal event pairs, determine a first causal relationship between a first element among the target elements of the one group and a second element of the target object, The unstructured text set can provide comments on the target object, Determining the target elements includes using a keyword extraction algorithm, Extract a plurality of keywords from the unstructured text set, Group at least some of the plurality of keywords based on the meanings of the plurality of keywords, Determining the target elements corresponding to the keywords based on the result of the grouping, An electronic device configured as such.

8. The second element Includes at least one of an element other than the first element in the target elements of the one group, a first predefined element of the target object, and an element of interest of the target object, The first predefined element is an element whose measurement criterion has a predetermined option, The element of interest is an aspect in which the target object is focused, The electronic device according to claim 7.

9. The unstructured text set consists of answers to open-ended questions in an information collection sheet for the target object, The information collection sheet includes closed-ended questions for the first predefined element, The electronic device according to claim 8.

10. Based on an unstructured text set regarding a target object of interest to a user, determine target elements of one group of the target object; and By analyzing the text in the unstructured text set, determine causal event pairs including cause events and result events; Determining a first causal relationship between a first element among the target elements of the one group and a second element of the target object based on the causal event pair; comprising; the unstructured text set can provide comments on the target object; determining the target element uses a keyword extraction algorithm; extracting a plurality of keywords from the unstructured text set; grouping at least a part of the plurality of keywords based on the meanings of the plurality of keywords; determining the target element corresponding to the keyword based on the result of the grouping, including; A computer-readable memory storing a computer program executable by a processor to implement a method of information processing.

Images