Angela Veyusmin Jehola Pengarang SEBASTIANUS ADI SANTOSO MOLA Dosen Pembimbing 1 MEITON BORU Dosen Pembimbing 2 Sebastianus Adi Santoso Mola Ketua Penguji Derwin Rony Sina Penguji 2 Meiton Boru Penguji 1

Social media does not only act as a means of communication to convey one's thoughts and feelings to others. The problem is that in written communications on social media, typos are often found. The existence of typos in addition to failing to adjust the position of the fingers with the keys on the keyboard, Typing errors (typo) can not only change the meaning of a word, The source of data in this study was taken from the Indonesian-English mixed word dataset obtained from the social media twitter. While the type of data used is qualitative data in the form of language data with non-standard words (NSW) and data from the Big Indonesian Language Dictionary (KBBI). To obtain the required data, the authors collect data through the twitter API. This study was tested using data of 4,378 non-standard words taken from the social media twitter. The test results obtained the highest MRR of 67.46% at Q = 9 where Q is a query with an average MRR of 43.59% for all ranks of Q and the initial letters with the highest values are "r", "c", "j" and "i". When compared to research (Lakonawa, 2020), there is an increase in the maximum MRR value of 15.13% and an increase in the average accuracy of 16.67%. This improvement shows that the position-based dynamic missmatch weighting method improves the accuracy of nonstandard word normalization with Nazief-Adriani stemming and Needleman-Wunsch string matching. Because the results of this study are not perfect, it is necessary to further develop dynamic weighting for other keyboard layouts. It is hoped that further research can be expanded by using other methods such as Jaro-Winkler Distance String Matching.