Linking Related Entities by Textually Described References

“Pushkin Digital” project is dedicated to structuring, grouping, and associating various objects including autographs, poems, prosaic an other types of texts written by A. S. Pushkin as well as scientific comments to these objects. The project is addressed to the problem of identifying and linking autographs references to the autographs themselves. Each autograph is encoded with simplified archival code. Solving this problem would allow to associate manuscripts with works they represent, for example sketches or drafts of an author's work. The solution we propose locates these codes in text of a scientific comment and linking it with the item in database. However, these codes can not be classified as parts of speech, which precludes the use of automatic text segmentation methods. To address this issue, an algorithm for text annotation was developed based on knowledge of the manuscript code structure. In this paper, we overview modern methods of named entity recognition in specific domains, explain chosen methodology for this problem, describe developed method and evaluate achieved results on the used dataset.