随着食品安全法规的不断发展,紧跟时事,及时了解条款修订情况对企业确保遵守并降低潜在的风险至关重要。利用计算机技术有助于自动识别监管变更,精简监察流程并能够及时地响应。该研究旨在探讨机器学习算法在食品安全风险管理中的应用。研究提出了一个提高合规风险管理效率和有效性的框架,基于Transformer的双向编码器表示(bidirectional encoder representations from transformers, BERT)——一个预训练的自然语言处理模型,以典型的监督学习模型为基线,被用来识别与特定食品类别潜在食品安全风险相关的新闻信息变化情况。新闻报道中的组织、食物、风险、法规等关键实体被BERT模型自动提取。以酒精饮料为例,结合领域专家提供的标注数据,研究得到了一个微调的(有提高的)BERT模型,该模型可以自动检测与酒精饮料和与之相关的关键实体相关的潜在监管变化。结果表明,相关性预测的F1分值为0.88,实体识别的F1分值为0.60。所提出的方法有可能显著减少手工工作,提高检测监管变化的准确性,最终强化食品企业的合规策略。
With the ever-evolving landscape of food safety regulations,staying abreast of modifications is crucial for enterprises to ensure adherence and mitigate potential risks.Leveraging computational technologies facilitates the automatic identification of regulatory changes,streamlining the monitoring process and enabling timely responses.This study explored the implementation of advanced machine learning algorithms in the context of the food industry,presenting a framework that enhanced the efficiency and effectiveness of compliance risk management.In this work,bidirectional encoder representations from transformers (BERT),a pretrained natural language processing model,was employed to identify regulatory news relevant to potential food safety risks of specific food categories,with typical supervised learning models as baselines.Key entities such as organization, food, contaminant,and regulation reported in the news reports were also automatically extracted by the BERT model.Using alcoholic beverage as an example along with the labeled data provided by domain experts,we obtained a fine-tuned BERT model that can automatically detect potential regulatory change related to alcoholic beverage and the critical entities associated with it.The results showed that the F1-score of relevance prediction was 0.81,and the F1-score of entity detection was 0.60.The proposed approach holds the potential to significantly reduce manual efforts,enhance accuracy in detecting regulatory alterations,and ultimately fortify the compliance strategies of food enterprises.
[1] HENSON S, CASWELL J.Food safety regulation:An overview of contemporary issues[J].Food Policy, 1999, 24(6):589-603.
[2] KOTSANOPOULOS K V, ARVANITOYANNIS I S.The role of auditing, food safety, and food quality standards in the food industry:A review[J].Comprehensive Reviews in Food Science and Food Safety, 2017, 16(5):760-775.
[3] MENSAH L D, JULIEN D.Implementation of food safety management systems in the UK[J].Food Control, 2011, 22(8):1216-1225.
[4] NGUYEN T T B, LI D.A systematic literature review of food safety management system implementation in global supply chains[J].British Food Journal, 2022, 124(10):3014-3031.
[5] HIGNETTE G, BUCHE P, COUVERT O, et al.Semantic annotation of Web data applied to risk in food[J].International Journal of Food Microbiology, 2008, 128(1):174-180.
[6] TAO D D, ZHANG D Y, HU R F, et al.Crowdsourcing and machine learning approaches for extracting entities indicating potential foodborne outbreaks from social media[J].Scientific Reports, 2021, 11(1):21678.
[7] WANG J, YUE H L.Food safety pre-warning system based on data mining for a sustainable food supply chain[J].Food Control, 2017, 73:223-229.
[8] NYCHAS G J E, PANAGOU E Z, MOHAREB F.Novel approaches for food safety management and communication[J].Current Opinion in Food Science, 2016, 12:13-20.
[9] WU Y N, CHEN J S.Food safety monitoring and surveillance in China:Past, present and future[J].Food Control, 2018, 90:429-439.
[10] PANG X N, LI Z J, CHEN J Y, et al.A comprehensive review of spirit drink safety standards and regulations from an international perspective[J].Journal of Food Protection, 2017, 80(3):431-442.
[11] 胡康, 王雅洁, 杨冰, 等.白酒质量安全风险分析与防范[J].中国酿造, 2019, 38(8):216-223.
HU K, WANG Y J, YANG B, et al.Risk analysis and prevention of Baijiu quality safety[J].China Brewing, 2019, 38(8):216-223.
[12] HE N X, BAYEN S.An overview of chemical contaminants and other undesirable chemicals in alcoholic beverages and strategies for analysis[J].Comprehensive Reviews in Food Science and Food Safety, 2020, 19(6):3916-3950.
[13] BŁASZCZYK I.The management of food safety in beverage industry[M]//Safety Issues in Beverage Production.Amsterdam:Elsevier, 2020:1-38.
[14] WANG Z F, QU P F, ZHAO Y F, et al.A nationwide survey and risk assessment of ethyl carbamate exposure due to daily intake of alcoholic beverages in the Chinese general population[J].Foods, 2023, 12(16):3129.
[15] WALDNER C.Big data for infectious diseases surveillance and the potential contribution to the investigation of foodborne disease in Canada[D].Winnipeg, Canada:National Collaborating Centre for Infectious Diseases, 2018.
[16] VAN DE BRUG F J, LUCAS LUIJCKX N B, CNOSSEN H J, et al.Early signals for emerging food safety risks:From past cases to future identification[J].Food Control, 2014, 39:75-86.
[17] MAEDA Y, KURITA N, IKEDA S.An early warning support system for food safety risks[M]//New Frontiers in Artificial Intelligence.Berlin, Heidelberg:Springer Berlin Heidelberg, 2006:446-457.
[18] KATE K, CHAUDHARI S, PRAPANCA A, et al.FoodSIS:A text mining system to improve the state of food safety in Singapore[C]//Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining.ACM, 2014:1709-1718.
[19] CHEN S Q, HUANG D D, NONG W Y, et al.Development of a food safety information database for Greater China[J].Food Control, 2016, 65:54-62.
[20] ZHU X Y, HUANG I Y, MANNING L.The role of media reporting in food safety governance in China:A dairy case study[J].Food Control, 2019, 96:165-179.
[21] BOUZEMBRAK Y, STEEN B, NESLO R, et al.Development of food fraud media monitoring system based on text mining[J].Food Control, 2018, 93:283-296.
[22] MARVIN H J P, HOENDERDAAL W, GAVAI A K, et al.Global media as an early warning tool for food fraud;an assessment of MedISys-FF[J].Food Control, 2022, 137:108961.
[23] TAO D D, YANG P K, FENG H.Utilization of text mining as a big data analysis tool for food science and nutrition[J].Comprehensive Reviews in Food Science and Food Safety, 2020, 19(2):875-894.
[24] ZHAI C X, MASSUNG S.Text Data Management and Analysis:A Practical Introduction to Information Retrieval and Text Mining[M]. New York:Association for Computing Machinery, 2016.
[25] OLDROYD R A, MORRIS M A, BIRKIN M.Identifying methods for monitoring foodborne illness:Review of existing public health surveillance techniques[J].JMIR Public Health and Surveillance, 2018, 4(2):e57.
[26] HU R, ZHANG D, TAO D, et al.TWEET-FID:An annotated dataset for multiple foodborne illness detection tasks[J].arXiv preprint arXiv, 2022, 2205.10726.
[27] QIAN C, MURPHY S I, ORSI R H, et al.How can AI help improve food safety?[J].Annual Review of Food Science and Technology, 2023, 14:517-538.
[28] ROBERTSON A S, REISIN MILLER A, DOLZ F.Supporting a data-driven approach to regulatory intelligence[J].Nature Reviews.Drug Discovery, 2021, 20(3):161-162.
[29] ROBERTS K, THAKKAR R, ALJUBURI L, et al.A vision for integrated publicly available information on regulated medical products[J].Clinical and Translational Science, 2022, 15(6):1321-1327.
[30] PATIL R S, KULKARNI S B, GAIKWAD V L.Artificial intelligence in pharmaceutical regulatory affairs[J].Drug Discovery Today, 2023, 28(9):103700.
[31] LI F Y, DONG S Y, LEIER A, et al.Positive-unlabeled learning in bioinformatics and computational biology:A brief review[J].Briefings in Bioinformatics, 2022, 23(1):bbab461.
