Using Natural Language Processing to Monitor Product Claims Compliance for FDA

DESCRIPTION (provided by applicant): Linguastat, Inc. proposes to develop a means to automate the process of monitoring and identifying companies engaged in false advertising and deceptive practices in the marketing of drugs, dietary supplements, and/or fo od products. By leveraging state of the art approaches in computational linguistics such as Information Extraction and Natural Language Processing, it should be feasible, with some adaptation, to use this technology to: 1) automatically and continuously mo nitor the websites, TV transcripts, press releases and other electronic marketing text communications of tens of thousands of companies for various claims and product information 2) automatically red-flag instances in which claims have a high likelihood of potential harm to consumers, according to FDA priorities 3) automatically identify and extract the companies, products and claims embedded in electronic product information and electronic promotional materials to create a database easily searchable by t he FDA and 4) automatically capture web-based or other electronic content for human review and store it as evidence. Such automated technology would enable the FDA to significantly stretch its limited human resource to more effectively and comprehensivel y identify noncompliant product information, detect deceptive ads and other illegal practices, successfully prosecute offenders, and prevent harm to American consumers. For this Phase I SBIR project we propose to assess the feasibility of automated claims monitoring in three steps: In the first step, we will train information extraction and natural language processing algorithms to extract product marketing claims from text. In the second, step we will apply data mining and rules-based algorithms to assess which claims are likely to be non-compliant and merit further attention by FDA staff. In the third step, we will design and build a database of product claims that allows analysts to search, organize, and prioritize product claims based on the type of clai m (e.g. what ailments does the product claim to treat), the type of product, and the likelihood of non- compliance. This technology will enable regulators and consumers to better monitor and detect cases of false, misleading, or deceptive advertising and p roduct information. By enabling more effective enforcement of FDA regulations and giving consumers tools to make better buying decisions, the public health can be better protected by minimizing the impact of products that cause harm, give false hope, or en tice consumers to forgo conventional remedies.