Text Data Negation and Similarity Measurement for Misinformation Management
Abstract Various methods are used to identify misinformation, and significant progress has been made. The methods usually notify users whether the information is true or fake, and not much else is provided. Other than the true or false answer, users may want more functions to be given. Some of the functions could be as follows. If the information is found to be fake, then what is the true information? How could the true information override the misinformation, so the mistake would not be made next time? In addition, is there a way to find out that the result is a false alarm? Simply providing the true or false answer for misinformation identification does not close the loop. This research discusses how to find the true information in the aftermath of misinformation detection by using several methods of text data processing.

This research takes the following steps after the misinformation is identified:

1. Negate the misinformation by using various NLP methods.
2. Search and rank the negated-misinformation by using a text similarity measurement.
3. Provide the true information to the user.

Misinformation research mostly shows whether the information is true or fake, but this does not close the loop because users may want more functions. This research tries to close the loop by providing the true information to the users. Other than finding the true information, a complete misinformation identification and management may include more functions like false-alarm processing and misinformation replacement.

Keywords
Security, misinformation, misinformation identification, NLP, natural language processing, text negation and antonym, text similarity measurement

References

1. (link) Verifying negative sentences
2. (link) Text data augmentations: Permutation, antonyms and negation
3. (link) Pragmatics and negative sentence processing
4. (link) Measuring sentences similarity: A survey

Conferences