Machine learning tools to improve the quality of imperfect keywords

The availability of keywords that describe the content of a text or document certainly is essential for effective and efficient content-based retrieval. But their quality, the presence of spelling variants, synonyms, near-synonyms, and spelling errors make their use less effective. Here we present a set of tools we are developing for the management of tags. These tools are intended to be used to improve the quality of textual features and to enhance traditional ways of searching and browsing data on the web. This approach integrates different methods: word embedding models, able to capture the semantics of words and their context, clustering algorithms, able to identify/group semantically related terms, and methods able to calculate the syntactic similarity between strings. The work is still under development, and the paper will present some preliminary qualitative results that demonstrate the feasibility of our approach.