Multi-modal language data processing method for non-universal language wisdom education

The invention provides a multi-modal language data processing method for non-universal language wisdom education, which comprises the following steps of: firstly, acquiring corpus information of a non-universal language through corpora disclosed by various universities and non-universal language data information disclosed by TED, Wikipedia and Opensubtitles webpages, and carrying out multi-modal data processing to obtain multi-modal data information of the non-universal language; the multi-modal data processing comprises voice processing of non-universal languages, Chinese translation processing, sentence pairing processing, synonym processing and similar-form word processing; secondly, using a key value pair to form a reverse index, when the reverse index is generated, generating mapping of index attributes and attribute related files, and storing the mapping as a plain text file; then storing the plain text file and the Chinese equivalent comparison line according to a fixed interval, and generating a corre