Topical and Non-Topical Approaches to Measure Similarity between Arabic Questions

Questions are crucial expressions in any language. Many Natural Language Processing (NLP) or Understanding (NLU) applications, such as question-answering computer systems, automatic chatting apps (chatbots), digital virtual assistants, and opinion mining, can benefit from accurately identifying similar questions an effective manner. We detail methods for similarities between Arabic that have been posted online by Internet users organizations. Our novel approach uses a non-topical rule-based methodology topical information (textual similarity, lexical semantic similarity) to determine if pair of similarly paraphrased. method counts the linguistic distances each question. Additionally, it identifies accordance with their format scope using expert hypotheses (rules) experimentally shown be useful practical. Even there is high degree similarity When question (Timex Factoid—inquiring about time) Who inquiry (Enamex Factoid—asking named entity), they will not similar. In experiment 2200 pairs, our attained accuracy 0.85, which remarkable given simplicity solution fact we did employ language models word embedding. order cover common queries presented users, gathered various forums resources. this study, describe unique detecting does require intensive processing, sizable corpus, costly repository. Because many rich textual resources, especially important informal text processing on Internet.