Abstract: In some implementations, a system is capable of using natural language processing (NLP), machine learning, and classification techniques to improve the storage and retrieval of text segments within a structured text repository. Data indicating a text segment retrieved from an online resource is retrieved. Metadata associated with the online resource is obtained. One or more one or more semantic scores are determined for the text segment. One or more classification labels are assigned to the text segment based at least on the obtained metadata and the one or more determined semantic scores. The text segment is indexed within a structured text repository using the one or more assigned classification labels.