Journal Recommendation System for Author Using Thai and English Information from Manuscript

There are thousands of academic journals in various fields of study. An article author must spend significant time searching and selecting a journal suitable for the article's content before submitting it to a journal for consideration. Since many articles are submitted to a journal at a time, it would take time for an editor to review, submit it to reviewers, and inform the results back to the author. If the article is rejected due to mismatched journal content, the author will spend more time to re-submit the article to another journal. Therefore, this research introduced a recommendation system to help the author choose an appropriate journal more effectively, based on TCI Thai Journals Online Database (ThaiJO). Data from Thai and English articles were used for analysis in this research. Our work involved studying the applied data, cleaning the data, and modeling, which included calculating the importance of text by Term Frequency - Inverse Document Frequency (TF-IDF), calculating similarity scores between articles and journals using Cosine Similarity and then ranking the scores to recommend the most suitable journal. This research experiment with modeling between a model from Thai data, a model from English data, and a model using both languages. The experiment shows that when we combine Thai and English keywords and abstract data, the accuracy in the form of hit rate is improved to 0.80650 from applying only English (0.78793) or Thai data (0.62888).