Feature location is a method of mapping between high abstraction of software artifacts into specific areas of source code. A use case scenario is a software artifact that contains many words that express the interaction between the actor and the system. The software is designed based on the story. Using a use case scenario as a query for the location of a feature would be beneficial, but not every word of a use case scenario can indicate the location feature. This study aims to measure the use of all words, only nouns, or nouns and verbs from the use case scenario as a query could help to find the features location. Natural language processing is also used as a preprocessing method to reduce the number of words in both the source code and the queries. The source code was indexed with LDA (Latent Dirichlet Allocation) to produce the topics. The topics proportion among the queries and source code were measure by cosine similarity ranking to identify that the queries similar with source code. A precision and recall were used to measure the successful rate of this method. The results are as follows, the average of precision of noun words was better (precision average 11.2%) than all word usage (precision average 9.9%). The average recall of noun word was better (recall average 4.2%) than all word usage (recall average 3.4%). The most interesting part was the noun words usage could minimize the area of documents by 83% (from 1074 documents into 181 documents) if compared with all words usage. The usage of Verb or nouns words in use case scenario got the best average recall (5.4%).