Method for extracting conceptual words from video subtitles

Abstract

The invention discloses a method for extracting conceptual words from video subtitles, which comprises the following steps of: carrying out word segmentation processing on a subtitle text, and deleting punctuation marks; stop words and part-of-speech tagging are carried out on the caption text after word segmentation; calculating co-occurrence characteristics of the target word and the adjacent word; calculating the semantic similarity between the target word and the adjacent word; performing concept word marking on a small number of subtitle texts subjected to word segmentation to serve as atraining set; and training a pre-established semi-supervised learning framework based on a conditional random field according to the training set to obtain a conceptual word prediction model, and obtaining a conceptual word prediction result corresponding to the subtitle text output by the conceptual word prediction model. Based on the method for extracting the conceptual words provided by the invention, the workload of manually labeling corpora is reduced, the accuracy of extracting the conceptual words in the MOOC video subtitle scene is improved, and the actual requirements are met.

Description

technical field [0001] The invention relates to a method for extracting concept words, in particular to a method for extracting concept words from video subtitles. Background technique [0002] Massive Open Online Courses (Massive Open Online Courses, referred to as MOOCs) have promoted knowledge sharing around the world due to their high-quality course resources, creating a large number of opportunities for teaching and learning in different disciplines. One of the basic steps in mining and analyzing MOOC platform data is to extract concept words in video subtitles. [0003] At present, although there are many methods to extract concept words from text, such as supervised and unsupervised methods based on various machine learning algorithms such as support vector machines and neural networks, supervised methods require a large amount of artificially labeled corpus, while unsupervised methods There is no need to manually annotate the corpus, but it is difficult to achieve s...

AI Technical Summary

Problems solved by technology

[0003] At present, although there are many methods to extract concept words from text, such as supervised and unsupervised methods based on various machine learning algorithms such as support vector machines and neural networks, supervised methods require a large amount of artificially labeled corpus, while unsupervised methods There is no need to manually annotate the corpus, but it is difficult to achieve satisfactory performance, and these methods are aimed at extracting keywords from general text mining scenarios. If they are directly applied to course video subtitle texts, they will usually not achieve satisfactory results. performance, this is because video subtitles are different from general text mining scenarios, such as academic papers and news texts, and keywords and concept words in the usual sense cannot be completely equated

The above aspects have brought certain difficulties to the concept word extraction task in video subtitles, so it is necessary to improve the existing keyword extraction methods

Images