Consider the following natural language information retrieval query: “I’m interested in articles about subject A but not related to subject B or subject C.” Is it possible to interpret a query like this in different ways? Common sense says that the answer to this question is yes and that the ambiguity stems from the use of the word “not” in the query. However, to support this intuitive answer, one may want to survey people.
The authors of this paper surveyed 64 people using 20 natural language queries that involved negation. They present the survey and its results, and implement a “fairly simple prototype” to interpret such queries. Their system provides possible groupings of relevant lexical units (that is, various interpretations of the queries). They state this and the paper ends. It ends where it should start.
Under the title of this paper, there are several questions to consider, including: How can we process queries with negations in natural language–based information retrieval systems? What is the effect of different approaches on system performance? and What is the significance of this problem in operational systems? The work is incomplete.