An improved system for identifying and compacting text data to be transmitted over communications lines and thereby reducing the data volume and transmission time. Transmitting and receiving text processing systems are provided identical library memories containing words commonly used in correspondence. Each word in a document to be communicated is compared to the transmitting system's word library and, if found in the library, only the library address is transmitted. If the word is not found in the library, then it is added to the transmitting system's library, sent, and added to the receiving system's library. The receiving system reconstructs the document by using the received addresses to access the appropriate words from its library and place them in the document. The system combines this word match encoding with character match encoding and facsimile run length encoding for communicating words not found in the system library. The character match process requires a template match and non-linear difference code summation combined with N-dimensional weighting using prestored feature vectors for statistically determining the match between an input character and characters stored in the system library.