This project will develop an artificial intelligence (AI)-based system to identify inconsistencies in traffic crash reports by analyzing crash narratives. Work in Stage 1 will focus on developing the proposed system. To maximize the developed system’s potential for practical applications, identification of crash factors essential for analysis and decision making will be prioritized, as well as those that frequently suffer from inconsistencies through literature review and traffic safety practitioners from state DOTs. A literature review of recent AI-NLP techniques will be conducted to ensure that the proposed tool incorporates the most recent and significant developments in the field. To analyze crash narratives, most promising NLP techniques will be implemented. The computational complexity of each technique/model will be examined to determine which techniques offer a convenient balance in capabilities and complexity for practical application. Work in Stage 2 will focus on the validation of the developed system as well as conducting the transfer to practice activities. A validation dataset will be developed based on manual human annotations. The validation will be two-fold. First, the system’s output for multiple crash factors will be compared against the results of manual identification performed by human annotators. Second, collaborating highway agencies will be asked to examine the system output and provide feedback on the relevance of results and functionalities that could enhance the system’s value for practical application. A series of transfer to practice activities will be initiated aimed at making the project findings known to a large audience of traffic safety practitioners. These activities include seeking commercialization opportunities by identifying potential licensees, hosting webinar for traffic safety professionals, disseminating the proposed system at transportation and traffic safety conferences. The final report will present research results, guidelines on the use of the system, and all relevant information, including the development of the NLP models, the data annotation process, the evaluation of the models, and the validation results.