Research On Internet Text Information Acquisition And Analysis Methods For Metallurgical Industry

The metallurgical industry plays an essential role in the national economy.It is essential to grasp and understand the development trends of the metallurgical industry in time.With the explosive growth of information in the metallurgical industry on the Internet,manual methods of collecting,sorting,and analyzing industry information can no longer meet the application needs.Therefore,we propose to study internet information collection and analysis methods for the metallurgical industry to improve the analysis efficiency and intelligence level.The research has an excellent supporting role in promoting the competitiveness of enterprises.Internet information acquisition and analysis for the metallurgical industry faces many key and challenging problems.The primary problem faced by the research is how to accurately identify metallurgical industry news from the massive pages of the Internet.Secondly,automatically classifying news from different sources into unified topic categories lays a foundation for large-scale,classified storage and query of industry news.In addition,how to automatically extract metallurgical industry knowledge from news texts and realize efficient and accurate industry news retrieval and recommendation.In this paper,we study these critical problems and propose corresponding solutions.1.This paper proposes a deep learning text classification model to identify metallurgical industry news from massive Internet news automatically.We construct a metallurgical industry recognition model by combining a pre-training language model and mean prototype network to alleviate the over-fitting problem of language model fine-tuning.The proposed method improves metallurgical industry news filtering performance.It provides a feasible technical scheme for obtaining large-scale metallurgical industry news.2.This paper proposes a metallurgical industry topic classification model based on prior Mixup data argumentation.It tackles the data imbalance problem by generating pseudo-samples in the text embedding space.The proposed method effectively improves the classification effect of news topics in the metallurgical industry.The proposed model provides a practical and feasible solution for implementing classified storage and query large-scale metallurgical industry news.3.This paper proposes a distantly supervised entity recognition model,which models nested and non-nested entities in a single model.The proposed method improves the negative sampling progress by inducting domain lexicon,which alleviates the influence of unlabeled entities on model training,improving entity recognition performance in the metallurgical industry.The proposed model can recognize enterprise names,product names,technical terms,and locations from metallurgical industry news.It lays a foundation for news retrieval and related news recommendation of the metallurgical industry based on entity keywords.4.This paper proposes a domain knowledge-aware document semantic hashing method.This method improves the semantic encoding process of documents by integrating metallurgical lexicon knowledge and nearest neighbor vocabulary information.The proposed model combines neighborhood component analysis and comparative learning,which improves the accuracy of similar news retrieval in the metallurgical industry.The proposed method solves the key problem of news similarity calculation in the large-scale metallurgical industry and provides a solution for fast similar news recommendations.Based on the proposed method and model,this paper constructs a prototype system of Internet information acquisition and analysis in the metallurgical industry.The system can facilitate the automatic acquisition and intelligent analysis of large-scale industry news.The developed functions of news classification query,entity keyword news retrieval,related news recommendation,and similar news recommendation verify the effectiveness of the proposed method and model from a practical view.