Social media short text named entity identification method based on visual object guidance

Abstract

The invention discloses a social media short text named entity identification method based on visual object guidance, which comprises the following steps of: establishing a text description data set of a social media short text, and providing visual image data related to text description; performing object detection on the visual image data, and performing visual word meaning extraction on the detected visual object category; carrying out semantic extraction on the original text description, and carrying out text feature extraction on text semantics; calculating the similarity between visual word meanings and text features, and performing weighted summation on the visual word meanings; calculating the correlation between the multi-modal semantics and the original text features to obtain afinal word meaning vector; and inputting the final word meaning vector into a conditional random field, and identifying entity boundaries and entity categories to obtain an identified named entity. According to the method, the entity type can be judged by utilizing the visual object information, and the social media short text data lacking semantic information is supplemented.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to a method for recognizing short text named entities in social media based on visual object guidance. Background technique [0002] Named entity recognition is a basic problem in natural language processing, and its purpose is to locate named entities from text segments and classify named entities into predefined categories (such as person names, place names, or organization names). Early named entity recognition methods were based on traditional machine learning methods, which required a large number of manually designed features (such as word frequency, word order, etc.). There are also some methods that use external resources (such as external dictionaries) to extract named entities. These methods often depend on specific task scenarios or the effectiveness of external resources, and the effect is poor in many specific scenarios. [0003] In recent years, d...

AI Technical Summary

Problems solved by technology

[0004] The word representation method that combines character features and word embedding models can effectively identify slang and some polysemous words in texts in the field of social media, but it performs poorly on short text datasets due to lack of semantics

Images