Author Indexing in Mathematics

238 p.Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1981.The American Mathematical Society (AMS) has required since 1970 that authors of articles published in AMS journals submit indexing terms with their manuscripts. This study examined 159 documents published by the AMS in 1975. Each contained author indexing terms--class numbers selected by authors from the AMS(MOS) Subject Classification Scheme (1970). Most documents also received indexing terms from professional indexers--Mathemaical Reviews editors. To compare these two indexing treatments, the investigator solicited readers--mathematics faculty and graduate students--to suggest queries to which the documents would be satisfactory responses. The investigator transformed the queries into search statements composed of AMS class numbers, and then he compared search terms with author and with professional indexing terms. A document was retrieved (and recall successful) if at least one search term matched at least one indexing term, under a given indexing treatment.The major hypothesis of the study was that authors would demonstrate better retrieval performance than professionals by having a higher recall score than the professionals. The results failed to support this hypothesis. Instead, there was no significant difference between authors and professionals: each group recalled 69 of the 138 documents that had been assigned a query and both types of indexing.Readers also assigned indexing terms to documents, although no reader assigned indexing terms and suggested a query for the same document. The investigator hypothesized that readers would have a higher recall score than professionals. The results failed to support this. Readers did have a higher recall score than professionals (68 of 123 documents retrieved compared to 64 of 123), but the higher reader score was not significant.An alternative to recall tests as a means to compare authors and professionals was the interindexer consistency (IIC) measure. The investigator hypothesized that author/reader IIC would be higher than either author/professional IIC or professional/reader IIC. The IIC measure was the number of terms common to two indexing treatments divided by the total number of unique terms in the two treatments. The results did not support the hypothesis. Author/reader IIC had a mean of 0.26 per document for 144 documents, while author/professional IIC was 0.42 for 157 documents, and professional/reader IIC was 0.28 for 142 documents. The lower author/reader IIC, compared to author/professional IIC, was significant, and the lower author/reader IIC, compared to professional/reader IIC, was not significant.The availability of both recall and interindexer consistency scores permitted a test of correlation. The investigator hypothesized a slight positive relationship between recall and interindexer consistency. The results supported this. Among 139 documents with at least two types of indexing, the Eta correlation value was 0.33; among 123 documents with all three types of indexing, the Eta correlation was 0.37. Both Eta values were significant.The investigator concluded that authors performed as well as professional indexers in this experiment, despite the assumed advantage of experience held by professionals. The failure to recall half the documents in the major hypothesis test was attributed in part to unfamiliarity of the investigator/searcher with specific subject areas in mathematics and to the failure of the AMS Scheme to have better systems of cross references, scope notes, hierarchical arrangement, and synonym control.U of I OnlyRestricted to the U of I community idenfinitely during batch ingest of legacy ETD