Method, search method and device for generating visual content based on text content

Abstract

The present invention relates to a method, a search method and a device for generating visual content based on text content. The method for generating personalized visual content based on text content includes the following steps: extracting information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation; converting the extracted dimensional information into visual features according to preset conversion rules; based on the quantitative representation of the dimensional information, quantizing the corresponding visual features according to a preset quantization algorithm to obtain the quantitative representation of the visual features; and combining multiple visual features together according to a combination rule, and modifying the quantitative attribute parameter value of the visual features based on the quantitative representation of each visual feature to obtain the visual content corresponding to the text content to be processed. The present invention provides a new content representation method and search method, increases the expressiveness of the content, and solves the content search problem in specific application scenarios.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, a search method and a device for generating visual content based on text content. Background Art

[0002] In most scenarios in various fields such as production and life, people are usually accustomed to using text content to express and communicate. With the development of visual technology and computer technology, more and more content can be presented in the form of images, audio and video. In some fields, visual content can quickly leave a deep impression on people and enhance people's understanding of the information to be conveyed. At present, high-quality visual content usually requires professionals to design and produce according to the content to be expressed, which is costly and time-consuming, and is not suitable for non-professionals to produce visual content in daily applications. Summary of the invention

[0003] In view of the technical problems existing in the prior art, the present invention proposes a method, a search method and a device for generating visual content based on text content, so as to provide a new content presentation method and search method.

[0004] According to one aspect of the present invention, the present invention provides a method for generating personalized visual content based on text content, comprising the following steps:

[0005] Extract information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation;

[0006] Convert the extracted dimensional information into visual features according to preset conversion rules;

[0007] Based on the quantized representation of the dimensional information, quantizing the corresponding visual feature according to a preset quantization algorithm to obtain a quantized representation of the visual feature; and

[0008] A plurality of visual features are combined together according to a combination rule, and a quantitative attribute parameter value of each visual feature is modified based on a quantitative representation of the visual feature to obtain visual content corresponding to the text content to be processed.

[0009] According to another aspect of the present invention, the present invention also provides a device for generating personalized visual content based on text content, comprising a text processing module, a first conversion module, a second conversion module and a feature combination module, wherein the text processing module is configured to extract information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation; the first conversion module is connected to the text processing module, and is configured to convert the extracted dimensional information into visual features according to preset conversion rules; the second conversion module is connected to the first conversion module, and is configured to quantize the corresponding visual features according to a preset quantization algorithm based on the quantitative representation of the dimensional information to obtain the quantitative representation of the visual features; the feature combination module is connected to the first conversion module and the second conversion module, and is configured to combine the multiple visual features obtained by the first conversion module, and modify the attribute parameter value of the visual feature based on the quantitative representation of each visual feature obtained by the second conversion module to obtain the visual content corresponding to the text content to be processed.

[0010] According to another aspect of the present invention, the present invention also provides a text search method, which comprises the following steps:

[0011] According to the aforementioned method for generating personalized visual content based on text content, corresponding target visual content is generated based on the search text content;

[0012] Performing a matching search in a visual content library based on the target visual content;

[0013] Obtaining a preset number of visual contents to be selected in descending order of matching degree with the target visual content; and

[0014] An original text for generating the visual content to be selected is determined based on the visual content to be selected.

[0015] According to another aspect of the present invention, the present invention also provides a text search device, which includes a visual content generation module, a content matching module and a text content determination module, wherein the visual content generation module is configured to generate corresponding target visual content based on the search text content according to the aforementioned method for generating personalized visual content based on text content; the content matching module is configured to perform a matching search in a visual content library based on the target visual content, and obtain a preset number of candidate visual contents in descending order of matching degree with the target visual content; and the text content determination module is configured to determine the original text of the candidate visual content based on the candidate visual content.

[0016] According to another aspect of the present invention, the present invention also provides an electronic device, which includes a processor and a memory, wherein a computer program instruction set is stored on the memory, and when the processor executes the computer program instruction set on the memory, the aforementioned method of generating personalized visual content based on text content or the aforementioned text search method is implemented.

[0017] According to another aspect of the present invention, the present invention also provides a computer-readable storage medium, wherein a computer program instruction set is stored on the computer-readable storage medium, and when the computer program instructions are executed by a processor, the aforementioned method for generating personalized visual content based on text content or the aforementioned text search method is implemented.

[0018] According to another aspect of the present invention, the present invention further provides a computer program product, which includes computer program instructions, and when the computer program instructions are executed by a processor, the aforementioned method for generating personalized visual content based on text content or the aforementioned text search method is implemented.

[0019] The present invention can generate corresponding visual content based on various text contents, change the content expression form from text to visual, and can increase the expressiveness of the content in specific application scenarios, or improve the efficiency of obtaining text expression content; on the other hand, the present invention also solves the content search problem in specific application scenarios through the text search method provided, and the search results have a high matching degree and fast speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The preferred embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, wherein:

[0021] Figure 1 is a flow chart of a method for generating personalized visual content based on text content according to Embodiment 1 of the present invention;

[0022] Figure 2 is a graphic representing resume text according to one embodiment of the present invention;

[0023] Figure 3 According to another embodiment of the present invention, Figure 2 A graphic representing the resume text based on the graphic shown;

[0024] Figure 4 is a flow chart of a method for generating personalized visual content based on text content according to Embodiment 2 of the present invention;

[0025] Figure 5 is a schematic diagram of an image expressing visual content obtained by the method described in Embodiment 2 of the present invention;

[0026] Figure 6is another image schematic diagram expressing visual content obtained by the method described in Embodiment 2 of the present invention;

[0027] Figure 7 is a flow chart of a method for generating personalized visual content based on text content according to Embodiment 3 of the present invention;

[0028] Figure 8 is a flow chart of a method for generating personalized visual content based on text content according to Embodiment 4 of the present invention;

[0029] Fig. 9 It is based on the user's Figure 5 The visual content shown is a schematic diagram of an image obtained by modifying the color of each polygon;

[0030] Fig.10 is a principle block diagram of a device for generating personalized visual content based on text content according to an embodiment of the present invention;

[0031] Fig.11 is a principle block diagram of a device for generating personalized visual content based on text content according to another embodiment of the present invention;

[0032] Fig.12 is a principle block diagram of a text-to-image conversion system according to an embodiment of the present invention;

[0033] Fig.13 is a principle block diagram of an HR management system according to an embodiment of the present invention;

[0034] Fig.14 is a flow chart of a content search method according to an embodiment of the present invention;

[0035] Fig.15 is a principle block diagram of a text search device according to an embodiment of the present invention; and

[0036] Fig.16 It is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] In the following detailed description, reference may be made to the various specification drawings that are part of the present application and are used to illustrate specific embodiments of the present application. In the accompanying drawings, similar reference numerals describe substantially similar components in different figures. The various specific embodiments of the present application are described below in sufficient detail so that a person of ordinary skill in the art with relevant knowledge and skills in the art can implement the technical solutions of the present application. It should be understood that other embodiments may also be used or structural, logical or electrical changes may be made to the embodiments of the present application.

[0039] Embodiment 1

[0040] Figure 1 is a flow chart of a method for generating personalized visual content based on text content according to Embodiment 1 of the present invention. In this embodiment, the method comprises the following steps:

[0041] Step S11, extracting information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation.

[0042] Step S12, converting the extracted dimensional information into visual features.

[0043] Step S13: quantizing the corresponding visual feature based on the quantized representation of the dimension information to obtain the quantized representation of the visual feature.

[0044] Step S14, combining multiple visual features together, and modifying the attribute value of each visual feature based on the quantitative representation of the visual feature, so as to obtain a visual file expressing the same content as the text to be processed.

[0045] Among them, in step S11, when extracting information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation, it can include multiple methods, for example, using a language model to perform semantic understanding on the content of the text to be processed, and the prompt information (Prompt) of the language model is preset with corresponding dimensional parameters, which include specific dimensional data, such as dimension names, and also include quantitative instructions. After the text to be processed is input into the language model, the language model performs semantic understanding on the content of the text to be processed according to the dimensional parameters preset in the Prompt, determines the specific content of the corresponding dimensional information, and according to the quantization instructions, uses the corresponding quantization algorithm to perform quantization according to the specific content of the dimensional information to obtain the quantitative representation of the dimensional information, and the quantitative representation is, for example, a score, a grade, etc.

[0046] The language model used in the aforementioned method is, for example, various existing open source pre-trained natural language models, such as the GPT series models, the Bert model, etc., which are obtained by fine-tuning these models and training them with suitable data.

[0047] For example, for a resume text, the dimension parameters specified in Prompt are "educational background", "work experience", "skills and expertise", "personality" and "professional image", and the quantitative data of "educational background" is specified to include four levels: "below undergraduate", "undergraduate", "graduate student" and "doctorate", and the corresponding numbers are 1-4. "Personality" is quantified from 0-100 points from introversion to extroversion. When a resume text is input into the language model, the language model outputs the specific content and corresponding quantitative values ​​of the five dimensions of "educational background", "work experience", "skills and expertise", "personality" and "professional image".

[0048] In another embodiment, in step S11, different dimension parameters are preset based on the difference of the text to be processed, and the dimension parameters include dimension data and corresponding quantitative data. Therefore, when extracting information from the content of the text to be processed, the dimension data is first read from the preset dimension parameters, and the content of the text to be processed is first classified, and the content is extracted, etc. to obtain the dimension information, for example, using a text classification algorithm combining traditional expert rules (Pattern), artificial feature engineering and shallow classification models, a text classification model based on deep learning, etc. Then, a quantitative representation is obtained based on other deep learning models or other quantitative rules according to the quantitative data.

[0049] For example, corresponding to the resume text, dimension parameters are preset, and the dimension data in the dimension parameters include "educational background", "work experience", "skills and specialties", "personality" and "professional image". The quantitative data in the dimension parameters are scores or levels corresponding to the dimension data, such as the quantitative data of "educational background" includes numbers 1-4, corresponding to the four levels of "undergraduate", "undergraduate", "graduate student" and "doctor". The quantitative data of "work experience" includes numbers 1-4, corresponding to "no experience", "1-3 years", "3-5" and "5-10". The quantitative data corresponding to "skills and specialties" are, for example, the levels of different types of skills, and are represented by specific numbers respectively. For example, corresponding to "programming" in the IT field, the quantitative data includes numbers 1-4, corresponding to the levels of "primary", "intermediate", "senior" and "expert", and the quantitative data for "personality" and "professional image" are scores of 0-100.

[0050] When it is determined that the text to be processed is a resume text, the five dimensional data of "education background", "work experience", "skills and expertise", "personality color" and "professional image" are read from the preset dimension parameters respectively, and then the corresponding dimensional information is extracted from the resume text content, and then the corresponding quantitative values ​​are determined from the dimensional information based on the respective quantitative data one by one. Different dimensional data have different algorithms for obtaining quantitative representation. Some quantitative representations can be obtained directly from the dimensional information. For example, when the resume text includes specific academic content (such as graduate students), the corresponding number can be determined as 3 based on the record of academic content (graduate students), that is, the quantitative representation of the dimension "education background" is obtained. However, the quantitative representation of some dimensions needs to be predicted by machine learning models, deep learning models, etc. For example, the two dimensions of "personality color" and "professional image" need to extract relevant content from the resume, and then predict it by the trained model to obtain the corresponding score. The quantitative representation of some dimensions can be obtained by statistics, calculating the average score, etc.

[0051] The visual features in step S12 are, for example, points, lines or two-dimensional graphics, or colors, or three-dimensional graphics, or animation clips, or animation units. The present invention collects the contents that can be used as visual features to form a visual feature library, and the aforementioned visual features have category attributes, combination attributes or other quantifiable attributes. The category attributes are, for example, "two-dimensional graphics", "three-dimensional graphics / models", "animation units", "animation clips", etc. The combination attributes refer to attributes that can be combined with their visual features. For example, the edges and vertices of two-dimensional polygons are respectively a combination attribute, and the faces and points in three-dimensional graphics can be respectively used as a combination attribute. The time between two images or two frames of "animation units" and "animation clips" can be used as a combination attribute. Other quantifiable attributes such as "RGB value of color", "contrast", "brightness", "length", "width", "area", "duration", etc., for certain categories of visual features also include unique attributes, for example, for two-dimensional polygons, its attributes include "number of sides", for sectors, its attributes include "center angle", "radius", "arc length", etc. Determined visual features can be obtained by setting attribute parameter values.

[0052] In one embodiment, the conversion rule described in step S12 includes conversion parameters, and a certain conversion rule can be obtained by setting the conversion parameters. For example, the visual feature category and the rules for selecting specific visual features are set in the conversion parameters. For example, in one embodiment, it is set to replace new visual feature categories according to a certain period; sort the visual features according to the storage address, select them from front to back, or randomly select them; the number of visual features that can be selected at one time can also be set, etc. For example, for the dimensions of "educational background", "work experience", "skills and expertise", "personality color" and "professional image" obtained from the resume text and their specific quantitative representation, a graphic can be randomly selected from the visual features of the category "two-dimensional graphics", such as selecting a sector as a visual feature.

[0053] After the visual features are determined in step S12, in step S13, the corresponding visual features are quantified according to the quantitative representation of the dimensional information to obtain the quantitative representation of the visual features. For example, after selecting the sector as the visual feature, the quantitative attributes of the sector that can be quantified include "center angle", "radius" and "arc length", and "center angle" is preferred. Therefore, in order to enable the sector to accurately represent the corresponding dimensional information, in one embodiment, the center angle of the circle with the vertex of the sector as the center is divided into five equal parts (corresponding to five dimensional information), and the center angle of each sector is 360 / 5=72 degrees as the full score value of a dimensional information. The corresponding center angle is determined according to the ratio of the specific quantization value of the dimension to the full score, so that a sector that accurately corresponds to it can be determined.

[0054] The combination rule in step S14 includes one or more combination parameters, and the corresponding combination rule can be obtained by setting the combination parameters. For example, the same combination attribute is set in one combination parameter for combination. After the visual features are determined, the specific combination attributes are set in another combination parameter. For example, after the fan-shaped visual features are determined, the vertices are specified for combining the various visual features. When combining, five sectors representing five dimensional information are combined together with the vertices of the sectors as the center of the circle. Thus, a graphic representing the resume text is obtained. Figure 2 In this embodiment, when the sectors are quantified, the quantification standard of the sectors corresponding to each dimension information is to quantify the central angle, so the distinction and difference of the five dimensional information can be conveniently displayed in the obtained graph, so that the user can easily determine which of the five aspects of the resume he / she has more advantages in.

[0055] Furthermore, in order to increase the readability of the visual content, text of the dimension information and / or symbols expressing the quantitative representation may be added to the visual content. The text or symbols may be added to the graphics, such as Figure 3 shown.

[0056] When the visual content is expressed by other types of visual features, the text of the dimensional information and / or the symbol expressing the quantitative representation may also be identified in the form of a legend.

[0057] Embodiment 2

[0058] Figure 4 is a flow chart of a method for generating personalized visual content based on text content according to Embodiment 2 of the present invention. In this embodiment, the method comprises the following steps:

[0059] Step S21, extracting information from the content of the text to be processed to obtain multiple dimensional information;

[0060] Step S22, further extracting multiple information elements from each dimension information;

[0061] Step S23, quantizing the information element based on the quantization standard of the information element to obtain a quantized representation of the information element;

[0062] Step S24, converting the extracted dimensional information into visual features.

[0063] Step S25, determining a visual feature attribute that matches the information element of the corresponding dimensional information;

[0064] Step S26, obtaining the attribute value of the visual feature based on the quantitative representation of the information element; and

[0065] Step S27, combining multiple visual features together, and modifying corresponding visual feature attribute parameter values ​​based on attribute values ​​corresponding to information elements, thereby obtaining a visual file expressing the same content as the text to be processed.

[0066] In this embodiment, after the dimension information is obtained in step S21, multiple information elements are further extracted for each dimension information in step S22. Still taking the resume text of the aforementioned embodiment one as an example, after the dimension information of "educational background", "work experience", "skills and expertise", "personality" and "professional image" is obtained, information is further extracted for each dimension information, wherein "educational background" obtains three information elements of "educational background", "school" and "grades", "work experience" obtains three information elements of "scale", "position" and "ability", and the other three dimensional information also obtain multiple information elements, which will not be repeated here. In step S23, the process of quantizing the information elements based on the quantization standard of the information elements to obtain the quantitative representation of the information elements is the same as the processing method of quantizing the dimension information in embodiment one. When implementing, you can refer to the description of quantizing the dimension information in embodiment one, which will not be repeated here.

[0067] In step S24, when the conversion rule is to select a two-dimensional graphic in order, a visual feature whose storage address is after the visual feature selected when the visual content was generated last time, such as a polygon, is selected.

[0068] In step S25, after determining the visual features corresponding to the dimensional information, determine the visual feature attributes that match the information element of the corresponding dimensional information, that is, use an attribute of the visual feature of the dimensional information that can express the visual content to express the information element. Taking a polygon as an example, the distance length from one vertex of the polygon to other vertices can be used as an information element. Taking a three-dimensional ball combination as an example, the different volumes or quantities of balls are used as an information element. When quantifying the visual features, the distance length between two vertices is determined based on the ratio of the score of the information element to the full score, or the number of three-dimensional balls is determined based on the score of the information element.

[0069] See also Figure 5 , which is a schematic diagram of an image expressing visual content obtained by the method described in Example 2 of the present invention. In this embodiment, a vertex of a polygon is used as a combination attribute to combine polygons corresponding to various dimensional information. In the polygon corresponding to each dimensional information, the distance length from the vertex used for combination to other vertices is expressed by the quantized value of an information element. In this embodiment, the quantized value of each information element has a full score of 10 points. Based on the score of each information element, the distance from the common vertex to the corresponding vertex is obtained. By connecting the vertices of a dimensional information, Figure 5 The graphic shown.

[0070] Furthermore, in order to simplify Figure 5 The generation process of the graph shown is based on the determined number of information elements, such as 15 in the figure. First, a center of a circle is determined, and then the center angle of the circle is evenly divided into 15 parts, and 15 points are obtained with the same radius (the full score of the dimension quantization value is 10). The 15 points are connected to obtain the base map. Based on the quantization value of each information element, the corresponding point is determined on the corresponding radius. The information element elements of the same dimension information are adjacent. Therefore, after determining the points corresponding to all information elements, the points corresponding to each information element and the center of the circle are connected to obtain Figure 5 That is to say, in this embodiment, points and lines are used as visual features corresponding to the dimension information. When combining, the combination rules are set to averagely diverge from the same center of the circle, and combination conditions are added to the combination rules, such as adding other auxiliary visual features, such as adding a base map, adding connecting lines of each line corresponding to the dimension information, and adding different colors. Various graphics can be obtained by setting the combination parameters in the combination rules.

[0071] For the convenience of observation, the name of the information element and the corresponding quantitative standard value are marked on the graph. In this embodiment, more text content is expressed by means of two-level information. In order to highlight the visual experience of different dimensional information, this embodiment also adds color as an attribute of dimensional information, and sets different colors for different dimensional information, thereby effectively distinguishing different dimensions, and intuitively expressing the content of the information element through the prominence of the vertices of the polygons constituting the dimension.

[0072] In order to highlight each visual feature obtained by combination, in another embodiment, auxiliary visual features can be added to the combined features, such as Figure 5 By adding a background map, the full score of the quantitative value and the difference between each information element and the full score can be more clearly displayed, so that users can understand the content expressed by the current information element more quickly and easily. Figure 5 All three information elements in the skill and expertise dimension are full marks, which shows that the skills of the current resume holder are outstanding. In the work experience dimension, the scale information element shows that the previous work unit of the current resume holder was relatively small.

[0073] The visual content in the above-mentioned embodiment is composed of two-dimensional graphics. When the selected visual feature is formed by a three-dimensional image, a three-dimensional visual content can be obtained, such as Figure 6 Each information element is represented by a three-dimensional balloon, and the quantized value of the information element is expressed by the height position of the three-dimensional balloon in the z direction. The larger the quantized value of the information element is, the higher its height in the z direction is.

[0074] In the aforementioned embodiments, in order to facilitate the comparison of dimensional information, the categories of visual features of each dimensional information are the same. However, in some application scenarios, they can also be different to highlight the differences between different dimensions. For example, in the field of advertising creativity, advertising images generated based on advertising texts can use different categories of visual features based on different dimensional information, such as visual content that combines animation and graphics. For another example, a unique virtual transportation card can be generated based on personal text content, etc.

[0075] In the various applications mentioned above, in order to increase the consistency in meaning between the visual content and the text content and highlight the extracted information, various auxiliary visual features can be added to the combination of visual features to strengthen the visual features corresponding to the dimensional information.

[0076] Embodiment 3

[0077] Figure 7 is a flow chart of a method for generating personalized visual content based on text content according to Embodiment 3 of the present invention. In this embodiment, the method comprises the following steps:

[0078] Step S31, extract information from the content of the text to be processed to obtain multiple dimensional information and determine the corresponding topic at the same time.

[0079] Step S32: quantize each dimension information to obtain a corresponding quantization value.

[0080] Step S33, converting the extracted dimensional information into visual features that conform to the theme according to preset conversion rules.

[0081] Step S34, combining multiple visual features according to the combination rule, and modifying the quantitative attribute parameter value of each visual feature based on the quantitative representation of the visual feature to obtain the visual content corresponding to the text content to be processed.

[0082] Compared with the aforementioned embodiment, this embodiment not only extracts information from the content of the text to be processed, but also determines the corresponding theme through semantic understanding. The theme can be referred to as the applicable scenario of the text content, such as the meaning of one or more combinations of formal, informal, commercial, academic, entertaining, relaxed, serious, funny, etc. In one embodiment, each visual feature stored in the visual feature library also includes a theme attribute, and one or more of the aforementioned theme contents are annotated in the theme attribute to determine the scenario in which the visual feature can be applied. For example, some animated image units composed of multiple funny expressions have the annotated content in their theme attributes such as "informal", "funny", "entertaining", and "relaxed".

[0083] This embodiment determines the scene that is applicable to the text content through semantic understanding, and then determines the visual features that match it, so that the generated visual content can be more consistent with the text content.

[0084] Embodiment 4

[0085] Figure 8 is a flow chart of a method for generating personalized visual content based on text content according to Embodiment 4 of the present invention. In this embodiment, the method comprises the following steps:

[0086] Step S41, using a large language model to extract information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation.

[0087] Step S42, input the obtained dimensional information and its quantitative representation into the fine-tuned and trained language model to generate a visual description text, wherein the visual features corresponding to the dimensional information and the quantitative representation of the visual features are specified in the visual description file.

[0088] Step S43, inputting the visual description text into the fine-tuned text graph model, the text graph model combines the visual features corresponding to the dimensional information specified in the visual description text according to the combination rule, and modifies the quantitative attribute parameter value of the visual feature based on the quantitative representation of each visual feature to obtain the visual content corresponding to the text content to be processed.

[0089] The processing process of step S41 is the same as that of the above-mentioned embodiment and will not be described again.

[0090] The large language model (LLM) in step S42 is, for example, an existing open source language model. Based on the requirements of the present invention, the existing open source language model is fine-tuned and trained using a training data set to make it more suitable for the present invention. For example: using GLM-10B, using a data set (a sample of the data set includes dimensional information and its quantification information, corresponding visual feature information and its quantification information, such as points and the number, size, and distribution of points, lines and line types, thickness, etc., and colors and their RGB values, etc.), efficient parameter fine-tuning is performed based on the strategy of P-tuning v2, so that the model can obtain a visual description text based on the input dimensional information and its quantification information, and specify the visual features corresponding to the dimensional information and the quantification representation of the visual features in the visual description text.

[0091] Similar open source language models include the GLM-130B model jointly trained by Tsinghua University KEG Laboratory and Zhipu AI; or Meta open source pre-trained language models, such as OPT-2.7B / 13B / 30B / 66B; or Bloom series models, such as Bloom-7B / 13B / 176B; or Open AI series models, GPT-4 / 3.5 / ; or Google's PaLM2 series models, such as Gecko, Otter, Buson, Unicorn, etc.

[0092] The Wensheng graph model in step S43 is, for example, DELL·E, DELL·E2, DELL·E3 models of Open AI, Parti and Imagen proposed by Google, Wenxin ERNIE-ViLG and Wenxin Yige models proposed by Baidu, and other models, such as CLIP+Disco Diffusion, Stable Diffusion model, etc. By calling any Wensheng graph model, the visual description text in step S43, different models obtain the corresponding image according to their preset algorithms, such as the combination rules set therein, image processing algorithms, such as image fusion, style transfer and other algorithms according to the visual features specified in the visual description text and the quantitative representation of the visual features.

[0093] Furthermore, when the expected visual content is a three-dimensional graphic, a moving picture or an animation, after obtaining the visualization description text, a description of generating the three-dimensional graphic, the moving picture or the animation is added thereto, and the parameters for generating the three-dimensional graphic and the parameters required for generating the moving picture are specified in the description, for example, the various surfaces and surface combination parameters constituting the three-dimensional graphic, or parameters including frame rate, duration, etc.

[0094] After obtaining the corresponding images, the aforementioned text-based graph model connects or splices the obtained images based on the aforementioned parameters to obtain corresponding three-dimensional graphics or animated graphics or animations.

[0095] After the visual content is obtained in each of the above embodiments, the visual content is stored as an image file in a static picture format or a dynamic picture format or a video file in a video format. For example, when the visual content is obtained in a two-dimensional graphic, it is stored in a commonly used picture format, such as JPEG, PNG, etc., and when the visual content is obtained in a two-dimensional graphic, it is stored in a format such as STL; when the visual content is obtained in a dynamic image unit, it is stored in a format such as GIF, WebP, APNG, etc., and when the visual content is obtained in an animation clip, a video clip, etc., it can be stored in a format such as WAV, MP4, MOV, etc.

[0096] Embodiment 5

[0097] In this embodiment, the visual feature has visual attributes, and the visual attributes can be category attributes, such as two-dimensional, three-dimensional, animated images, etc. The visual attributes can be colors, such as different color systems, or specific colors, etc. It can also be the theme attributes in the aforementioned embodiments, and can also be other quantitative attributes. After obtaining the visual content based on the aforementioned embodiments, if the user is not satisfied with the current visual content, the user can select the corresponding visual attribute parameters and change the attribute values. The visual attributes of the visual features specified in the update instruction are modified based on the visual attribute parameter values ​​in the update instruction issued by the user, thereby obtaining new visual content. Fig. 9 As shown, when the user Figure 5 The visual content shown is a schematic diagram of the image obtained by modifying the color of each polygon.

[0098] Embodiment 6

[0099] In this embodiment, after obtaining the visual content corresponding to the text content to be processed according to any of the above embodiments, in order to prevent the current visual content from being repeated, the current visual content is further checked in this implementation to see whether it is the same as the historical visual content. For example, the current visual content is compared with the historical visual content for similarity, and when the similarity is greater than a certain threshold, the two are considered to be the same. If the current visual content is the same as the historical visual content, a change operation is performed to make the latest visual content different from the historical visual content.

[0100] The change operations include, for example, the following:

[0101] The first one is to perform data backtracking, which goes back to the time when the extracted dimensional information is converted into visual features according to the preset conversion rules. When the conversion rule is to randomly select visual features in an attribute category, the current attribute category also includes multiple other visual features to be selected. In this way, the current visual features are changed to other visual features to be selected.

[0102] The second method is to perform data backtracking, backtracking to when the extracted dimensional information is converted into visual features according to the preset conversion rules, and changing the conversion parameter value in the conversion rule to obtain a new visual feature. For example, changing the category attribute specified in the conversion parameter value, and selecting a new visual feature in the new category attribute.

[0103] The third method is to change the combination parameter values ​​in the combination rules to obtain a new visual feature combination relationship.

[0104] Embodiment 7

[0105] In this embodiment, in order to meet the needs of users in various application scenarios, multiple corresponding visual contents are obtained by processing the text content to be processed. Referring to the above-mentioned embodiments 1 to 5, the purpose of obtaining multiple corresponding visual contents can be achieved by any of the following methods:

[0106] 1. The preset conversion rule includes a conversion quantity parameter, and the conversion quantity parameter value is set to be greater than 1, and each dimensional information is converted into multiple visual features determined by the conversion quantity parameter value, so that multiple corresponding visual contents can be obtained when the visual features are combined.

[0107] Second, the combination parameter values ​​in the combination rule are set to multiple values, and a visual feature combination relationship is obtained with each combination parameter value.

[0108] In another aspect, the present invention provides an apparatus for generating personalized visual content based on text content, see Fig.10 , Fig.10It is a principle block diagram of a device for generating personalized visual content based on text content according to an embodiment of the present invention. The device for generating personalized visual content based on text content in this embodiment (referred to as text-to-image conversion device 100) includes a text processing module 110, a first conversion module 120, a second conversion module 130 and a feature combination module 140, wherein the text processing module 110 is used to extract information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation. The first conversion module 120 is connected to the text processing module 110, and converts the extracted dimensional information into visual features according to preset conversion rules. The second conversion module 130 is connected to the first conversion module 120, and quantizes the corresponding visual features based on the quantitative representation of the dimensional information according to a preset quantization algorithm to obtain the quantitative representation of the visual features. When the feature combination module 140 is connected to the first conversion module 120 and the second conversion module 130 respectively, it is configured to combine the multiple visual features obtained by the first conversion module 120, and modify the attribute parameter value of the visual feature based on the quantitative representation of each visual feature obtained by the second conversion module 130 to obtain the visual content corresponding to the text content to be processed.

[0109] Fig.11 1 is a principle block diagram of a device for generating personalized visual content based on text content according to another embodiment of the present invention. The device for generating personalized visual content based on text content (referred to as text-to-image conversion device 100) in this embodiment includes a first language module 101, a second language module 102 and a text-to-image module 103, wherein the first language module 101 calls a first language model to extract information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation.

[0110] The second language module 102 is connected to the first language module 101, and inputs the extracted multiple dimensional information and quantitative representations into the fine-tuned second language model to generate a visual description text, wherein the visual features corresponding to the dimensional information and the quantitative representations of the visual features are specified in the visual description file.

[0111] The text graph module 103 inputs the visual description text into the fine-tuned text graph model, and the text graph model combines the visual features corresponding to the dimensional information specified in the visual description text according to the combination rule, and modifies the quantitative attribute parameter value of the visual feature based on the quantitative representation of each visual feature to obtain the visual content corresponding to the text content to be processed.

[0112] In the above two embodiments, before information extraction, the received data is processed into a text to be processed. Among them, the received data is, for example, text data or voice data. If it is voice data, the text data is obtained by voice recognition technology. For text data, firstly, speech text preprocessing is performed, mainly including removing punctuation marks, converting uppercase letters to lowercase letters, removing stop words, etc. Then, keyword extraction is performed, and the keyword extraction methods include methods based on preset dimensions, methods based on statistics, methods based on machine learning, and methods based on deep learning, etc. Then, semantic analysis is performed, and the semantic analysis methods, for example, use rule-based methods, statistics-based methods, and methods based on deep learning, etc. The text processing module 110 can use common semantic analysis models, such as Word2Vec, BERT, RoBERTa, etc. to implement the above data processing, and finally clean and integrate the data, remove duplicate information, fill in missing information, and unify information in different formats, etc. The first language module 101 can implement the above processing process through large language models (Large Language Models, referred to as LLM), general language models (General Language Model, referred to as GLM), etc.

[0113] Fig.12 1 is a block diagram of a text-to-image conversion system according to an embodiment of the present invention. The text-to-image conversion system 10 in this embodiment includes a text-to-image conversion device 100 and an interactive module 200. The text-to-image conversion device 100 and the interactive module 200 are respectively connected to a database 300 and a user information library 400. The interactive module 200 is used to provide an interactive interface for interacting with a user. The user can input text, set parameters, and view the obtained visual content through the interactive interface.

[0114] The user can directly input text through the input area of ​​the interactive interface, and the interactive module 200 sends the text data input in the input area to the text-to-image conversion device 100. Alternatively, the user inputs a voice in the input area of ​​the interactive interface by voice, and the interactive module 200 sends the voice data to the text-to-image conversion device 100. The user can also send a specified file to the text-to-image conversion device 100 by sending a file.

[0115] The user can set parameters through the interactive interface, such as setting one or more of the number, type (2D, 3D, animated image, animation, etc.), and theme (formal, entertainment, serious, relaxed, etc.) of generated visual content. The interactive module 200 sends these parameter data of the user device to the text-to-image conversion apparatus 100.

[0116] After receiving the text data, the text-to-image conversion device 100 processes the text data based on the parameter data set by the user. The processing process is the same as any of the aforementioned method embodiments, which will not be repeated here. The text-to-image conversion device 100 sends the obtained visual content in the form of data or files to the interactive module 200. After the interactive module 200 receives the visual content in the form of data, it displays it in the display area of ​​the interactive interface. After the interactive module 200 receives the visual content file, it displays a notification that the file has been generated in a pop-up window or the like in the interactive interface. The user can open the file according to the storage address in the notification to view the generated visual content.

[0117] The database 300 stores various contents that can be used as visual features, such as points, lines, two-dimensional graphics, three-dimensional graphics, animation clips, or animated image units, and stores various attribute data of each visual feature, such as category attributes, combination attributes, quantitative attributes, etc. The database 300 also stores other data required by the text-to-image conversion device 100 when generating visual content, such as prompts used by the language model, etc. There can be multiple prompts, each adapted to different application scenarios.

[0118] In one embodiment, a user becomes a user of the system by registering personal information with the text-to-image conversion system 10, and thus the text-to-image conversion system also maintains a user information library 400, in which a log of the user's application of the text-to-image conversion device 100 is maintained, so that each time visual content is generated based on the text input by the user, it is possible to check whether the current visual content is the same as the historical visual content. When the current visual content is the same as the historical visual content, the obtained visual content can be made the same as the historical visual content through a change operation.

[0119] In one embodiment, after the user applies the text-to-image conversion device 100 to obtain visual content, the interactive module 200 provides the user with modifiable attributes through an interactive interface, for example, in the form of a drop-down menu or a list, or based on the user's selection operation or click operation on the visual features constituting the visual content, the attribute options are provided. Figure 5 and Fig. 9 For example, when the user clicks on a polygon corresponding to a dimension information, a color attribute setting option will appear. After the user selects a color and confirms, the interactive module 200 obtains a corresponding update instruction based on the user operation and sends it to the text-to-image conversion device 100. The text-to-image conversion device 100 modifies the color of the corresponding polygon in the visual content based on the update instruction, thereby obtaining Fig. 9 In addition to changing the color, users can also change the theme, category, etc. of the visual content. I will not go into details here.

[0120] Fig.133 is a block diagram of the HR management system according to an embodiment of the present invention. In this embodiment, the HR management system 20 communicates with the aforementioned text-to-image conversion system 10. The database 300 stores job seekers' resumes. The HR management system 20 sends a conversion instruction to the text-to-image conversion system 10, and the text-to-image conversion system 10 generates corresponding visual content for each job seeker's resume, such as Figure 5 or Fig. 9 The two-dimensional image shown, or Figure 6 The HR management system 20 displays the image in a separate column when displaying a resume list. When HR browses the resume list of job seekers, they can understand the key content of the resume through the corresponding two-dimensional image displayed in the list without opening the resume, thereby saving HR time in browsing resumes and improving work efficiency.

[0121] Since the present invention can convert text content into visual content, and store the two-dimensional images, three-dimensional graphics / models, animated image units, animations and other contents converted from text and expressing the same meaning in the form of image files and video files, the present invention provides a novel text search method, see Fig.14 , Fig.14 This is a flow chart of a content search method according to an embodiment of the present invention. It includes the following steps:

[0122] Step S61, generating corresponding target visual content based on the search text content.

[0123] Step S62: performing a matching search in a visual content library based on the target visual content.

[0124] Step S63: obtaining a preset number of visual contents to be selected in descending order of matching degree with the target visual content.

[0125] Step S64: determining and generating the original text of the visual content to be selected based on the visual content to be selected.

[0126] In this embodiment, when the user needs to search for files that meet certain conditions from the database 300, a search text is created. Fig.13 Taking the HR management system shown in the figure as an example, when HR needs to screen out resumes that meet certain conditions, a search text is created in which the required conditions are described, such as skills, work experience, education, etc. In step S61, the corresponding visual content can be generated for the search text content through the methods provided by the various embodiments described above, such as Figure 5 or Figure 6 or Fig. 9The image shown is then used as the target in step S62 to perform a matching search in the image library corresponding to the resume in the database 300. The database 300 stores visual content, such as images, generated based on any of the aforementioned methods using the resume as text, and the resume and its image are made to correspond one to one through a set identifier (such as a file name). The matching search process in step S62, for example, calculates the similarity of the target image with the images in the image library one by one, and then sorts them from high to low according to the similarity (or matching degree). The image similarity calculation can be implemented in a variety of ways or algorithms, such as using cosine similarity, hash algorithm, histogram distance, structural similarity (such as measurement based on multiple indicators such as brightness, contrast, structure, etc.), pixel matching calculation and other algorithms for calculation, or based on a deep learning model obtained through training for calculation.

[0127] In step S63, one or more images with the highest ranking are selected as the images to be selected. The number of images to be selected is, for example, one or more, or a number set by the user. After the images to be selected are determined, in step S64, the corresponding resume is read from the database according to the one-to-one correspondence between text and image, and presented to the HR user.

[0128] Although the text search method is illustrated by taking a resume and its corresponding image as an example in this embodiment, the method is also applicable to other texts and corresponding three-dimensional graphics, animated images, and animations. In these scenarios, due to different visual contents, the matching search process of visual contents is slightly different. For example, for the similarity calculation of three-dimensional graphics, it is necessary to calculate the similarity of three dimensions; for animated images and animations, it is necessary to cut frames and then calculate the image similarity frame by frame or calculate the image similarity of key frames.

[0129] Correspondingly, the present invention also provides a text search device, see Fig.15 , Fig.15 It is a principle block diagram of a text search device according to an embodiment of the present invention, wherein the text search device 30 includes a visual content generation module 31, a content matching module 32 and a text content determination module 33, wherein the visual content generation module 31 generates corresponding target visual content based on the search text content according to the method for generating personalized visual content based on text content in any of the aforementioned embodiments. For example, the visual content generation module 31 can be the aforementioned text-to-image conversion device 100 or text-to-image conversion system 10. The content matching module 32 performs a matching search in the database 300 based on the target visual content, and obtains a preset number of selected visual contents in order of the degree of matching with the target visual content from high to low. The text content determination module 33 determines the original text of the selected visual content based on the selected visual content.

[0130] In an application embodiment, the text search device 30 is connected to the HR management system 20, for example. The HR user inputs the search text through the interface provided by the HR management system 20 and sends the search text to the text search device 30. After obtaining the matching resume text, the text search device 30 sends it to the HR management system 20, so that HR can search for resumes that meet its conditions and requirements from a large number of resume files.

[0131] In another application embodiment, the text-to-image conversion system 10 and the text search device 30 are integrated on a recruitment platform, and the recruitment platform includes a recruitment user terminal and a job-seeking user terminal. The recruitment platform generates corresponding recruitment visual content based on the recruitment information released by the recruiter and generates corresponding resume visual content based on the resume of the job seeker through the text-to-image conversion system 10, and stores them in the platform database. Recruiting users can search for resumes that meet their conditions and requirements through the text search device 30, and job-seeking users can search for recruitment information that meets their conditions and requirements through the text search device 30. In addition, the recruiting users and job-seeking users can generate corresponding images according to the text descriptions they provide through the text-to-image conversion system 10, and add them to the recruitment information or resume, thereby forming a unique visual experience.

[0132] Fig.16 1 is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. The electronic device can be implemented as a server or other various computing devices, such as a desktop personal computer, a tablet computer, a laptop computer, etc., including a processor 601 and a memory 602. The memory 602 stores a program instruction set. When the processor 601 executes the program instruction set on the memory 602, the aforementioned method for generating personalized visual content based on text content or the text search method is implemented.

[0133] Specifically, the processor 601 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiment of the present invention.

[0134] The memory 602 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 602 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 602 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 602 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 602 is a non-volatile solid-state memory.

[0135] The memory may include a read-only memory (ROM), a random access memory (RAM), a disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Therefore, generally, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the method for generating personalized visual content based on text content or the text search method provided by the present invention.

[0136] In one example, the electronic device may further include a communication interface 603 and a bus 604. The processor 601, the memory 602, and the communication interface 603 are connected via the bus 604 and communicate with each other.

[0137] The communication interface 603 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiment of the present invention.

[0138] Bus 604 includes hardware, software or both, and the components of online data flow billing equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 604 may include one or more buses. Although the embodiment of the present invention describes and shows a specific bus, the present invention considers any suitable bus or interconnection.

[0139] The present invention also provides a computer-readable storage medium having computer program instructions stored thereon, which can be executed by a processor to implement the method for generating personalized visual content based on text content or the text search method in the aforementioned embodiment. The computer-readable storage medium can be any medium that can tangibly contain or store computer-executable instructions for use by or in combination with an instruction execution system, device, and apparatus. The storage medium can be a transient computer-readable storage medium or a non-transient computer-readable storage medium. Non-transient computer-readable storage media may include, but are not limited to, magnetic storage devices, optical storage devices, and / or semiconductor storage devices. Examples of such storage devices include, for example, magnetic disks, optical disks based on CD, DVD, or Blu-ray technology, and persistent solid-state memories such as flash memory, solid-state drives, and the like.

[0140] The present invention also provides a computer program product, which includes computer program instructions, and when the computer program instructions are executed by a processor, the method for generating personalized visual content based on text content or the text search method in the above-mentioned embodiment is implemented. The computer program product includes but is not limited to application installation packages published on websites and application stores, application plug-ins, and small programs that can be run in certain applications.

[0141] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0142] The above embodiments are only used to illustrate the present invention, but not to limit the present invention. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the scope of the present invention. Therefore, all equivalent technical solutions should also fall within the scope of the present invention.

Claims

1. A method for generating personalized visual content based on text content, comprising: Extract information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation; Convert the extracted dimensional information into visual features according to preset conversion rules; Based on the quantitative representation of the dimensional information, quantizing the corresponding visual feature according to a preset quantization algorithm to obtain a quantitative representation of the visual feature; as well as Combining multiple visual features together according to a combination rule, and modifying the quantitative attribute parameter value of each visual feature based on the quantitative representation of each visual feature to obtain visual content corresponding to the text content to be processed; Wherein, the visual feature is a point, a line or a two-dimensional figure, or a color, or a three-dimensional figure, or an animation clip, or an animation unit, and the quantitative attribute parameter value is used to obtain the determined visual feature; Among them, further include: Performing semantic understanding on the content of the text to be processed, and obtaining a corresponding visual theme based on the semantic understanding; Correspondingly, when the extracted dimensional information is converted into visual features, visual features that match the visual theme are determined from the visual feature library, and after combining multiple visual features together, they conform to the visual theme. Each visual feature stored in the visual feature library also includes theme attributes, and the theme attributes are used to label one or more theme contents. The multiple visual features are converted from the multiple dimensional information.

2. The method for generating personalized visual content based on text content according to claim 1, wherein: When extracting multiple dimensional information and its quantitative representation, it further includes: Extract multiple information elements based on each dimension information; quantizing the information element based on a quantization standard of the information element to obtain a quantized representation of the information element; Correspondingly, after converting the extracted dimensional information into visual features, determining visual feature attributes that match the information elements of the corresponding dimensional information; Obtaining the attribute value of the visual feature based on the quantized representation of the information element; and The corresponding visual feature attribute parameter value is modified based on the obtained attribute value.

3. The method for generating personalized visual content based on text content according to claim 1, wherein: When the extracted dimensional information is converted into visual features, the visual features corresponding to each dimensional information have the same one or more combined attributes and their parameters; correspondingly, when multiple visual features are combined together, the parameter value of the combined attribute is set, and a combination relationship of multiple visual features is established based on the combined attribute parameter value, and the multiple visual features are obtained by converting the multiple dimensional information.

4. The method for generating personalized visual content based on text content according to claim 1, wherein: The visual features have category attributes, and the categories are points, lines or two-dimensional graphics, or colors, or three-dimensional graphics, or animation clips, or animated image units.

5. The method for generating personalized visual content based on text content according to claim 1, wherein: When the extracted dimensional information is converted into visual features, the category attributes of the visual features corresponding to each dimensional information are the same.

6. The method for generating personalized visual content based on text content according to claim 1, wherein: When the extracted dimensional information is converted into visual features, the quantitative attributes of the visual features corresponding to each dimensional information are the same.

7. The method for generating personalized visual content based on text content according to claim 1, wherein: After combining the multiple visual features, an auxiliary visual feature is added to the combined feature to obtain visual content corresponding to the text content to be processed.

8. The method for generating personalized visual content based on text content according to claim 1, wherein: Texts of dimension information and / or symbols expressing quantitative representation are added to the visual content.

9. The method for generating personalized visual content based on text content according to claim 1, wherein: The steps of extracting information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation include: Based on the language model, semantic understanding and quantitative analysis are performed on the content of the text to be processed, and multiple dimensional information and corresponding quantitative representations are extracted from the content of the text to be processed.

10. The method for generating personalized visual content based on text content according to claim 1, wherein: The steps of extracting information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation include: Reading dimension data from preset dimension parameters, and extracting information from the content of the text to be processed according to the dimension data to obtain corresponding dimension information; The dimension information is processed according to the quantized data in the dimension parameter to obtain a quantized representation of the dimension information.

11. The method for generating personalized visual content based on text content according to claim 1 or 10, wherein: Further including: Inputting the extracted multiple dimensional information and quantized representations to the fine-tuned language model to generate a visual description text, wherein the visual features corresponding to the dimensional information and the quantized representations of the visual features are specified in the visual description text; Correspondingly, the visual description text is input into the fine-tuned text graph model, which combines the visual features corresponding to the dimensional information specified in the visual description text according to the combination rules, and modifies the quantitative attribute parameter value of the visual feature based on the quantitative representation of each visual feature to obtain the visual content corresponding to the text content to be processed.

12. The method for generating personalized visual content based on text content according to claim 1, wherein: The visual feature has visual attributes. After obtaining the visual content corresponding to the text content to be processed, in response to the received visual attribute update instruction, the visual attribute of the visual feature specified in the update instruction is modified according to the visual attribute parameter value in the update instruction to obtain new visual content.

13. The method for generating personalized visual content based on text content according to claim 1, wherein: After obtaining the visual content corresponding to the text content to be processed, the method further includes: Check whether the current visual content is the same as the historical visual content; In response to the current visual content being the same as the historical visual content, performing one or more of the following change operations to make the latest visual content different from the historical visual content; The change operation includes: When converting the extracted dimensional information into visual features according to a preset conversion rule, when multiple visual features to be selected are obtained through the conversion, the current visual feature is changed to another visual feature to be selected; Changing the conversion parameter values ​​in the conversion rules to obtain new visual features; The combination parameter values ​​in the combination rule are changed to obtain a new visual feature combination relationship.

14. The method for generating personalized visual content based on text content according to claim 1, wherein: Based on one or more of the following operations to obtain multiple visual contents: A conversion quantity parameter is set in the preset conversion rule, and each dimensional information is converted into a plurality of visual features determined by the conversion quantity parameter value based on the conversion quantity parameter value; A plurality of combination parameter values ​​are set in the combination rule, and a visual feature combination relationship is obtained with each combination parameter value.

15. The method for generating personalized visual content based on text content according to claim 1, further comprising: The visual content is stored as an image file in a static picture format or a dynamic picture format or a video file in a video format.

16. A device for generating personalized visual content based on text content, comprising: A text processing module, configured to extract information from the content of the text to be processed to obtain multiple dimensional information and its quantitative representation; A first conversion module, connected to the text processing module, configured to convert the extracted dimension information into visual features according to a preset conversion rule; A second conversion module, connected to the first conversion module, is configured to quantize the corresponding visual features according to a preset quantization algorithm based on the quantization representation of the dimension information to obtain the quantization representation of the visual features; as well as a feature combination module, connected to the first conversion module and the second conversion module, configured to combine the multiple visual features obtained by the first conversion module, and modify the quantitative attribute parameter value of the visual feature based on the quantitative representation of each visual feature obtained by the second conversion module, so as to obtain visual content corresponding to the text content to be processed; Wherein, the visual feature is a point, a line or a two-dimensional figure, or a color, or a three-dimensional figure, or an animation clip, or an animation unit, and the quantitative attribute parameter value is used to obtain the determined visual feature; Wherein, the text processing module is further configured as follows: Performing semantic understanding on the content of the text to be processed, and obtaining a corresponding visual theme based on the semantic understanding; Correspondingly, when the extracted dimensional information is converted into visual features, visual features that match the visual theme are determined from the visual feature library, and after combining multiple visual features together, they conform to the visual theme. Each visual feature stored in the visual feature library also includes theme attributes, and the theme attributes are used to label one or more theme contents. The multiple visual features are converted from the multiple dimensional information.

17. A text search method, comprising: According to any of the methods of claims 1-15, the search text content is used as the text content to be processed, and the corresponding target visual content is generated based on the search text content; Performing a matching search in a visual content library based on the target visual content; Obtaining a preset number of visual contents to be selected in descending order of matching degree with the target visual content; as well as An original text for generating the visual content to be selected is determined based on the visual content to be selected.

18. A text search device, wherein: include: A visual content generation module, configured to use the search text content as the text content to be processed according to any one of the methods of claims 1-15, and to generate corresponding target visual content based on the search text content; a content matching module connected to the visual content generating module and configured to perform a matching search in a visual content library based on the target visual content, and obtain a preset number of visual contents to be selected in descending order of matching degree with the target visual content; as well as The text content determination module is connected to the content matching module and is configured to determine and generate the original text of the visual content to be selected based on the visual content to be selected.

19. An electronic device comprising a processor and a memory, wherein a computer program instruction set is stored in the memory, and when the processor executes the computer program instruction set in the memory, the method for generating personalized visual content based on text content as described in any one of claims 1 to 15 or the text search method as described in claim 17 is implemented.

20. A computer-readable storage medium, wherein: The computer-readable storage medium stores a set of computer program instructions, and when the computer program instructions are executed by the processor, the method for generating personalized visual content based on text content described in any one of claims 1-15 or the text search method described in claim 17 is implemented.

21. A computer program product, comprising computer program instructions, wherein when the computer program instructions are executed by a processor, the method for generating personalized visual content based on text content according to any one of claims 1 to 15 or the text search method according to claim 17 is implemented.

Images