Investigating the Impacts of Misspellings in Patent Search by Combining Natural Language Tools and Rule-Based Approaches

Among all sources of technical information, patent information is one the richest and most comprehensive. Knowing how to search in this mass documents becoming increasingly crucial. However, many users have limited knowledge patents strategies, so they must use intuitive, often approximate approaches that can lead highly inaccurate searches be time-consuming. To address problem, there are tools help expand queries increase recall as not miss good documents, however, it remains an open problem dealing with misspellings-based strategies. Typically, presence misspellings text underestimated even by experts field, no specific functionality handle available, both free paid. The goal article raise awareness about difficulties making a proper strategy also takes into account possible misspellings. It important know where we expect find them much these may affect final result. In particular, chosen divide categories, distinguishing between associated generic keyword or multiword from acronyms, chemical formulas, names applicants, inventors, formulas theorems. At least example case given for each category, showing when Finally, integrated approach combining word contextual embedding models based on deep learning rule-based algorithm wild cards truncation operators suggested correcting query, automatically suggesting consistent misspellings, thus achieving more accurate reliable