Short video editing method based on natural language

Abstract

The invention provides a short video editing method based on a natural language, and the method comprises the following steps: S1, building a video material database, and adding classification tags for video materials; s2, acquiring an initial video editing demand taking a natural language as an information carrier; s3, performing semantic analysis on the initial video editing demand, and matching the natural language information with the classification label according to a semantic matching rule to obtain a demand keyword list; s4, retrieving a video material database according to the demand keyword to obtain a video material; s5, the user confirms the video material; and S6, combining the video materials confirmed by the user to generate video data meeting a video editing requirement. The method has the beneficial effects that video editing is performed through natural language interaction, and the operation is simple and convenient; and the universality is good.

Description

Technical Field

[0001] The present invention relates to the technical field of video editing, and in particular to a natural language-based short video intelligent editing method. Background Art

[0002] Currently, short video editing is widely used in advertising, entertainment, education and other fields. The existing short video intelligent editing method needs to preset labels for video clips, and then combine the video clips together through text feature comparison and retrieval. The editing effect is highly dependent on the accuracy of the labels themselves, and there are problems of cumbersome operation and low efficiency. Especially when users need to combine multiple video clips and edit them with audio and text subtitles to produce personal video works, non-professional users usually find it difficult to express complete and accurate video information parameters in advance. Video editing work is like looking for a needle in a haystack and even has no direction. Therefore, a more intelligent and efficient video editing method is needed to meet the needs of selecting and editing massive videos. Summary of the invention

[0003] The present invention aims to provide a short video editing method based on natural language, which realizes efficient and convenient video editing through natural language interaction.

[0004] To achieve the above object, the present invention is implemented through the following technical solutions:

[0005] A short video editing method based on natural language comprises the following steps: S1 establishes a video material database and adds classification tags to the video materials; S2 obtains initial video editing requirements with natural language as the information carrier; S3 performs semantic analysis on the initial video editing requirements, matches the natural language information with the classification tags according to semantic matching rules, and obtains a demand keyword list; S4 retrieves the video material database according to the demand keywords to obtain the video material; S5 the user confirms the video material; S6 combines the video materials confirmed by the user to generate video data that meets the video editing requirements.

[0006] Furthermore, the classification labels in step S1 include primary labels: image material labels, audio material labels and text material labels; secondary labels: scene labels, emotion labels and object labels.

[0007] Furthermore, in step S2, natural language information is converted into text information through speech recognition to obtain initial video editing requirements.

[0008] Further, step S3 includes S301: performing word-by-word semantic extraction on the text information converted from natural language information; S302: performing semantic extraction on adjacent word permutations and combinations of the text information converted from natural language information; S303: generating a classification label list by judging the semantics of single words and word permutation and combination through semantic compatibility-mutual exclusion rules; S304: asking the user whether the labels in the classification label list need to be marked one by one; S305: generating a required keyword list containing several video classification labels after the user confirms.

[0009] Furthermore, in step S5, if the number of detected video materials is greater than the set value, the supplementary tag information of the detected video materials other than the required keyword list is extracted to ask questions to the user, and the user answers the questions to add supplementary tags. The detected video materials are searched again through the supplementary tags until the number of video materials is less than or equal to the set value; if the number of detected video materials is less than or equal to the set value, the corresponding video material is displayed and the user is prompted to select and confirm.

[0010] The present invention has the following beneficial effects: (1) video editing is performed through natural language interaction, which is simple and convenient to operate; (2) video materials are accurately screened by gradually guiding users to confirm, and a small number of optional video materials are displayed to users, avoiding searching for a needle in a haystack; (3) compatible semantic matching rules are adapted, and the marking rules of video material classification labels in different databases can be applied, with good versatility. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0012] In order to better understand the essence of the present invention, the technical solutions in the embodiments of the present invention are clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its applications.

[0013] Reference Figure 1 In this embodiment, a short video editing method based on natural language is provided, which is characterized by comprising the following steps: S1 establishes a video material database and adds classification labels to the video materials; S2 obtains the initial video editing requirements with natural language as the information carrier; S3 performs semantic analysis on the initial video editing requirements, matches the natural language information with the classification labels according to the semantic matching rules, and obtains a demand keyword list; S4 retrieves the video material database according to the demand keywords to obtain the video material; S5 the user confirms the video material; S6 combines the video materials confirmed by the user to generate video data that meets the video editing requirements.

[0014] The classification labels in step S1 include primary labels: image material labels, audio material labels and text material labels; secondary labels: scene labels, emotion labels and object labels. The methods for classifying video materials and labeling corresponding labels are relatively mature in the prior art, and many of them can be corrected and updated by collecting massive user feedback, so they will not be elaborated here.

[0015] In step S2, natural language information is converted into text information through speech recognition to obtain initial video editing requirements.

[0016] Step S3 includes S301: extracting word-by-word semantics from text information converted from natural language information; S302: extracting semantics from adjacent word permutations and combinations of text information converted from natural language information; S303: generating a classification label list by judging the semantic compatibility-mutual exclusion rule for the semantics of single words and word permutations and combinations; S304: asking the user whether the labels in the classification label list need to be marked one by one; S305: generating a demand keyword list containing several video classification labels after the user confirms. The semantic inclusion and exclusion rule is already relatively mature in the field of speech recognition and will not be described in detail.

[0017] In step S5, if the number of detected video materials is greater than the set value, the supplementary tag information of the detected video materials other than the required keyword list is extracted to ask questions to the user, and the user answers the questions to add supplementary tags. The detected video materials are searched again through the supplementary tags until the number of video materials is less than or equal to the set value; if the number of detected video materials is less than or equal to the set value, the corresponding video material is displayed and the user is prompted to select and confirm.

[0018] The user converts natural language into an initial demand keyword list, such as "The afternoon sun is bright, XX is active in the mountains". The system searches and extracts the video material that needs to be synthesized by asking questions such as the type, quantity, and specific vectorization parameters. The user answers and confirms in natural language, such as whether background audio, subtitles, and transition effects between video clips need to be added. Video editing is achieved conveniently and quickly through natural language interaction.

[0019] It should be noted that, for those skilled in the art, the embodiments of this specification may have various equivalent substitutions or changes. Any modification, equivalent substitution, improvement, etc. made within the scope of the working principle of the embodiments of the present invention shall be included in the protection scope of the present invention.

Claims

1. A short video editing method based on natural language, characterized in that The following steps are involved: S1 establishes a video material database and adds classification tags to the video materials; S2 obtains the initial video editing requirements with natural language as the information carrier; S3 performs semantic analysis on the initial video editing requirements, matches the natural language information with the classification tags according to the semantic matching rules, and obtains a list of required keywords; S4 retrieves the video material database according to the required keywords to obtain the video material; S5 the user confirms the video material; S6 combines the video materials confirmed by the user to generate video data that meets the video editing requirements.

2. The natural language-based short video intelligent editing method according to claim 1, characterized in that: The classification labels in step S1 include primary labels: image material labels, audio material labels and text material labels; secondary labels: scene labels, emotion labels and object labels.

3. The natural language-based short video intelligent editing method according to claim 1, characterized in that: In step S2, natural language information is converted into text information through speech recognition to obtain initial video editing requirements.

4. The natural language-based short video intelligent editing method according to claim 1, characterized in that: Step S3 includes S301: performing word-by-word semantic extraction on text information converted from natural language information; S302: performing semantic extraction on adjacent word permutations and combinations of text information converted from natural language information; S303: generating a classification label list by judging the semantics of single words and word permutation and combination semantics through semantic compatibility-mutual exclusion rules; S304: asking the user whether the labels in the classification label list need to be marked one by one; S305: generating a required keyword list containing several video classification labels after the user confirms.

5. The natural language-based short video intelligent editing method according to claim 1, characterized in that: In step S5, if the number of detected video materials is greater than the set value, the supplementary tag information of the detected video materials other than the required keyword list is extracted to ask questions to the user, and the user answers the questions to add supplementary tags. The detected video materials are searched again through the supplementary tags until the number of video materials is less than or equal to the set value; if the number of detected video materials is less than or equal to the set value, the corresponding video material is displayed and the user is prompted to select and confirm.

Images