Page 157 - JOURNAL OF LIBRARY SCIENCE IN CHINA 2018 Vol. 42
P. 157

156 Journal of Library Science in China, Vol. 8, 2016


              One-to-many comparisons are advantage and disadvantage comparisons between one
            method and many other ones. When compared with a family of methods, words like
            traditional, classic, previous, existing, usual, and ordinary are quite frequently used. When
            compared with several methods, coordinating conjunctions or punctuations that represent
            coordinating conjunctions are used, for example, compared with ... and ..., ... is more .... One-
            to-many comparison describes more of advantage. Usually, one-to-many comparison does
            not discuss disadvantage. In some one-to-many comparisons, there are specific objects to be
            compared while in some others, there is no explicit reference, for example, “compared with
            other methods, MCLP is more concise and applicable”. Under such a situation no method
            that is compared with can be extracted.



            3.4  Method process description

            Method process description includes processes, steps and conditions. There are many rules for
            method process description, some of the rules contain apparent feature words like first, then,
            etc. For example, “Yeung et al. designed an algorithm, first using clustering to label objects and
            tags to different topics, then user tag sets are investigated to determine the user tag vector under
            different interest topics”, “In this paper we use spectrum characteristics arrangement-based
            transfer classification method to predict customer defection. The process is as follows. First, we
            designed methods for data dimension fusion and data attribute normalization to normalize the
            data representation. Then we use spectrum characteristics arrangement to establish the mapping
            between different fields of data. This will achieve approximate unified characteristics distribution
            between heterogeneous domains. Finally, we use TSVM classification model for customer
            classification. Thus we can recognize the loyalty of customers and customer churn.
              Some method process descriptions describe presupposition as well as conditions. For example,
            “The presupposition of this method is that every author contributes the same to the paper, that is
            they have the same contribution factor”. Some descriptions just have no rules. For example, in the
            personalized sorting method we use the user search history information to train user interest model,
            and use collaborative recommendation algorithm to obtain neighbor users with common interests.
            We sort the search results according to the extent and degree of recommendation of neighbor users
            as well as the similarity between the user interest model and the document. In this description,
            there is no apparent indicator word, so it is difficult for us to recognize whether the sentence is
            about method process or not.
              Overall speaking, instead of using one sentence, method process description usually requires a
            sentence set or a paragraph. So it is difficult to construct method process rules.
   152   153   154   155   156   157   158   159   160   161   162