Similarity measures have been proven over time for their usefulness in solving real-time problems. Researchers modify and develop new methods to suit different areas of application relative to specific data sets. The existing similarity measures on some application domains gave ineffective results. This study modified existing similarity measures and converted them into intuitionistic fuzzy similarity measure for word cognition detection in dementia patients. Similarity measures of bigram, Dice and Canberra were modified as Dice1, Dice2, Canberra1, Canberra2 and bigram which were extended to intuitionistic fuzzy similarity measures, and the text and pattern were classified into type-1 linguistic variable. Experiments were conducted on both existing and modified methods with stored data sets while evaluations were performed on generated results using average similarity measure, average processing time and root mean square error. Experimental results indicated that modified Dice2 gave the highest similarity value of 0.98, followed by Dice, modified Dice1, modified Canberra1, modified Canberra2, Canberra, modified bigram with the values of 0.93, 093, 0.90, 0.89, 0.84 and 0.82 respectively for 100 pairs of text and pattern matching of equal length of characters. Considering the processing time for experimental cases of 20, 50 and 100 sets of text and pattern matching of both equal and unequal length of strings, modified Dice2 computed the lowest processing time followed by modified Dice1, modified bigram, Canberra, modified Canberra1, Dice, and modified Canberra2. MD2 gave a more effective and efficient IFSM compared to existing IFSM for dementia patients. The results of root mean square error on both existing and modified methods indicated that modified Dice2 has the lowest value. Modified Dice2 classified the word entered by the user as against word generated randomly by the computer as type-1 linguistic variable of simple, moderate and high as suggested by the experts. Modified Dice2 gave the highest similarity value, lowest processing time, lowest root mean square error which could be used for word cognition detection and hence an enhanced IFSM for dementia patients
